feat(installer): improve cursor and opencode skill handling

README.md

@@ -18,11 +18,17 @@ ai-coding-skills/
 ├── README.md
 ├── docs/
 │   ├── README.md
+│   ├── INSTALLER.md
+│   ├── CODEX.md
+│   ├── CLAUDE-CODE.md
+│   ├── CURSOR.md
+│   ├── OPENCODE.md
 │   ├── ATLASSIAN.md
 │   ├── CREATE-PLAN.md
 │   ├── DO-TASK.md
 │   ├── IMPLEMENT-PLAN.md
 │   ├── PI.md
+│   ├── PI-COMMON-REVIEWER.md
 │   ├── PI-RESEARCH.md
 │   ├── PI-SUPERPOWERS.md
 │   ├── TELEGRAM-NOTIFICATIONS.md
@@ -61,6 +67,7 @@ ai-coding-skills/
 │   └── web-automation/
 │       ├── codex/
 │       ├── claude-code/
+│       ├── cursor/
 │       ├── opencode/
 │       └── pi/
 ├── .codex/
@@ -98,20 +105,54 @@ ai-coding-skills/
 | implement-plan | pi | Worktree-isolated plan execution with iterative cross-model milestone review | Ready | [IMPLEMENT-PLAN](docs/IMPLEMENT-PLAN.md) |
 | web-automation | codex | CloakBrowser-backed browsing, scraping, auth, flow automation, and install validation | Ready | [WEB-AUTOMATION](docs/WEB-AUTOMATION.md) |
 | web-automation | claude-code | CloakBrowser-backed browsing, scraping, auth, flow automation, and install validation | Ready | [WEB-AUTOMATION](docs/WEB-AUTOMATION.md) |
+| web-automation | cursor | CloakBrowser-backed browsing, scraping, auth, flow automation, and install validation | Ready | [WEB-AUTOMATION](docs/WEB-AUTOMATION.md) |
 | web-automation | opencode | CloakBrowser-backed browsing, scraping, auth, flow automation, and install validation | Ready | [WEB-AUTOMATION](docs/WEB-AUTOMATION.md) |
 | web-automation | pi | CloakBrowser-backed browsing, scraping, auth, flow automation, and install validation | Ready | [WEB-AUTOMATION](docs/WEB-AUTOMATION.md) |
 
-- Docs index: `docs/README.md`
-- Atlassian guide: `docs/ATLASSIAN.md`
-- Create-plan guide: `docs/CREATE-PLAN.md`
-- Do-task guide: `docs/DO-TASK.md`
-- Implement-plan guide: `docs/IMPLEMENT-PLAN.md`
-- Web-automation guide: `docs/WEB-AUTOMATION.md`
+- Start with the docs index: `docs/README.md`
+- Automated install/update/remove wizard: `docs/INSTALLER.md`
+- Manual install by client: `docs/CODEX.md`, `docs/CLAUDE-CODE.md`, `docs/CURSOR.md`, `docs/OPENCODE.md`, `docs/PI.md`
+- Skill guides: `docs/ATLASSIAN.md`, `docs/CREATE-PLAN.md`, `docs/DO-TASK.md`, `docs/IMPLEMENT-PLAN.md`, `docs/WEB-AUTOMATION.md`
 
 ## Compatibility Policy
 
 Each skill should explicitly document agent compatibility and any prerequisites directly in its own `SKILL.md`.
 
+## Skill Manager Wizard
+
+Use the repository skill manager to install, update/reinstall, or remove skills for supported local clients:
+
+```bash
+./scripts/manage-skills.sh
+# or
+node scripts/manage-skills.mjs
+```
+
+Useful non-interactive modes and examples:
+
+```bash
+# Preview detected clients and planned changes without writing files
+node scripts/manage-skills.mjs --dry-run
+
+# Emit a machine-readable operation plan from an answers file
+node scripts/manage-skills.mjs --plan-only --answers answers.json
+
+# Install/update a skill for a specific client
+node scripts/manage-skills.mjs --client codex --scope global --skill create-plan --action install --yes
+node scripts/manage-skills.mjs --client codex --scope global --skill create-plan --action update --yes
+
+# Remove a skill from a specific client
+node scripts/manage-skills.mjs --client claude-code --scope global --skill do-task --action remove --yes
+
+# Install the Pi package globally or project-locally through the manager
+node scripts/manage-skills.mjs --client pi --scope packageGlobal --pi-package --action install --yes
+node scripts/manage-skills.mjs --client pi --scope packageLocal --pi-package --action install --yes
+```
+
+The wizard detects Codex, Claude Code, Cursor, OpenCode, and Pi, previews operations, checks Superpowers dependencies for workflow skills, and prints a final operation report.
+
+`ai_plan/` is gitignored local planning state used by `create-plan` and `do-task`. The skill manager does not install, sync, or publish `ai_plan/` contents.
+
 ## Pi Package
 
 The repo root now includes a pi package manifest that ships only the pi-specific surface:
@@ -119,8 +160,11 @@ The repo root now includes a pi package manifest that ships only the pi-specific
 - `pi-package/skills/*/` for the five packaged Pi skills
 - `skills/reviewer-runtime/pi/`
 - `docs/PI*.md`
+- `scripts/manage-skills.mjs` and `scripts/manage-skills.sh`
 - `scripts/sync-pi-package-skills.sh`
 - `scripts/verify-pi-resources.sh`
+- `scripts/verify-pi-workflows.sh`
+- `scripts/verify-reviewer-support.sh`
 
 Install it from a cloned checkout with the repo-owned one-liner:
 

docs/ATLASSIAN.md

@@ -2,7 +2,7 @@
 
 ## Purpose
 
-Provide a portable Atlassian Cloud skill for Codex, Claude Code, Cursor Agent, and OpenCode using one shared CLI surface for common Jira and Confluence workflows.
+Provide a portable Atlassian Cloud skill for Codex, Claude Code, Cursor Agent, OpenCode, and Pi using one shared CLI surface for common Jira and Confluence workflows.
 
 ## Why This Skill Exists
 
@@ -117,6 +117,28 @@ cd ~/.cursor/skills/atlassian/scripts
 pnpm install
 ```
 
+### Pi
+
+Recommended full Pi package install:
+
+```bash
+./scripts/install-pi-package.sh --global
+# or, for project-local Pi package install
+./scripts/install-pi-package.sh --local
+```
+
+Manual single-skill Pi install from the package mirror:
+
+```bash
+./scripts/sync-pi-package-skills.sh
+mkdir -p .pi/skills/atlassian
+cp -R pi-package/skills/atlassian/* .pi/skills/atlassian/
+cd .pi/skills/atlassian/scripts
+pnpm install --frozen-lockfile
+```
+
+Global manual installs use `~/.pi/agent/skills/atlassian/` instead of `.pi/skills/atlassian/`.
+
 ## Verify Installation
 
 Run in the installed `scripts/` folder:

docs/CLAUDE-CODE.md

@@ -0,0 +1,57 @@
+# Claude Code Manual Install
+
+## Skill Root
+
+Claude Code's Skill tool auto-discovers skills installed under:
+
+```bash
+~/.claude/skills/<skill-name>/
+```
+
+Manual install example:
+
+```bash
+mkdir -p ~/.claude/skills/do-task
+cp -R skills/do-task/claude-code/* ~/.claude/skills/do-task/
+```
+
+Use `skills/<skill>/claude-code/*` for each supported skill.
+
+## Reviewer Runtime
+
+```bash
+mkdir -p ~/.claude/skills/reviewer-runtime
+cp skills/reviewer-runtime/run-review.sh ~/.claude/skills/reviewer-runtime/
+cp skills/reviewer-runtime/notify-telegram.sh ~/.claude/skills/reviewer-runtime/
+chmod +x ~/.claude/skills/reviewer-runtime/*.sh
+```
+
+## Superpowers
+
+Workflow skills invoke Superpowers through Claude Code's skill system. Install or link the Obra Superpowers skills under:
+
+```bash
+~/.claude/skills/superpowers/
+```
+
+Example:
+
+```bash
+ln -s /absolute/path/to/obra/superpowers/skills ~/.claude/skills/superpowers
+```
+
+Verify:
+
+```bash
+test -f ~/.claude/skills/superpowers/brainstorming/SKILL.md
+test -f ~/.claude/skills/superpowers/test-driven-development/SKILL.md
+test -f ~/.claude/skills/superpowers/verification-before-completion/SKILL.md
+```
+
+## Verify
+
+```bash
+claude --version
+test -f ~/.claude/skills/do-task/SKILL.md
+test -x ~/.claude/skills/reviewer-runtime/run-review.sh
+```

docs/CODEX.md

@@ -0,0 +1,61 @@
+# Codex Manual Install
+
+## Skill Root
+
+Codex skills are installed under:
+
+```bash
+~/.codex/skills/<skill-name>/
+```
+
+Manual install example:
+
+```bash
+mkdir -p ~/.codex/skills/create-plan
+cp -R skills/create-plan/codex/* ~/.codex/skills/create-plan/
+```
+
+Repeat with `skills/<skill>/codex/*` for `atlassian`, `create-plan`, `do-task`, `implement-plan`, and `web-automation`.
+
+## Reviewer Runtime
+
+Workflow skills need the shared runtime helpers:
+
+```bash
+mkdir -p ~/.codex/skills/reviewer-runtime
+cp skills/reviewer-runtime/run-review.sh ~/.codex/skills/reviewer-runtime/
+cp skills/reviewer-runtime/notify-telegram.sh ~/.codex/skills/reviewer-runtime/
+chmod +x ~/.codex/skills/reviewer-runtime/*.sh
+```
+
+## Superpowers
+
+Workflow skills require Obra Superpowers. Codex variants expect the shared root:
+
+```bash
+mkdir -p ~/.agents/skills
+ln -s /absolute/path/to/obra/superpowers/skills ~/.agents/skills/superpowers
+```
+
+Verify planning skills:
+
+```bash
+test -f ~/.agents/skills/superpowers/brainstorming/SKILL.md
+test -f ~/.agents/skills/superpowers/writing-plans/SKILL.md
+```
+
+Verify execution skills:
+
+```bash
+test -f ~/.agents/skills/superpowers/test-driven-development/SKILL.md
+test -f ~/.agents/skills/superpowers/verification-before-completion/SKILL.md
+test -f ~/.agents/skills/superpowers/finishing-a-development-branch/SKILL.md
+```
+
+## Verify
+
+```bash
+test -f ~/.codex/skills/create-plan/SKILL.md
+test -x ~/.codex/skills/reviewer-runtime/run-review.sh
+codex --version
+```

docs/CREATE-PLAN.md

@@ -14,10 +14,11 @@ Create structured implementation plans with milestone and story tracking, and op
   - `~/.agents/skills/superpowers -> ~/.codex/superpowers/skills`
 - For Cursor, skills must be installed under `.cursor/skills/` (repo-local) or `~/.cursor/skills/` (global)
 - Shared reviewer runtime must be installed beside agent skills when using reviewer CLIs:
-  - Codex: `~/.codex/skills/reviewer-runtime/run-review.sh`
-  - Claude Code: `~/.claude/skills/reviewer-runtime/run-review.sh`
-  - OpenCode: `~/.config/opencode/skills/reviewer-runtime/run-review.sh`
-  - Cursor: `.cursor/skills/reviewer-runtime/run-review.sh` or `~/.cursor/skills/reviewer-runtime/run-review.sh`
+  - Codex: `~/.codex/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
+  - Claude Code: `~/.claude/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
+  - OpenCode: `~/.config/opencode/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
+  - Cursor: `.cursor/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}` or `~/.cursor/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
+  - Pi: `.pi/skills/reviewer-runtime/pi/{run-review.sh,notify-telegram.sh}` or `~/.pi/agent/skills/reviewer-runtime/pi/{run-review.sh,notify-telegram.sh}`
 - Telegram notification setup is documented in [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md)
 
 If dependencies are missing, stop and return:
@@ -33,6 +34,7 @@ To use the iterative plan review feature, one of these CLIs must be installed:
 | `codex` | `npm install -g @openai/codex` | `codex --version` |
 | `claude` | `npm install -g @anthropic-ai/claude-code` | `claude --version` |
 | `cursor` | `curl https://cursor.com/install -fsS \| bash` | `cursor-agent --version` (binary: `cursor-agent`; alias `cursor agent` also works) |
+| `pi` | Install Pi coding agent | `pi --version`; list models with `pi --list-models [search]` |
 
 The reviewer CLI is independent of which agent is running the planning — e.g., Claude Code can send plans to Codex for review, and vice versa.
 
@@ -87,6 +89,31 @@ mkdir -p ~/.cursor/skills/reviewer-runtime
 cp -R skills/reviewer-runtime/* ~/.cursor/skills/reviewer-runtime/
 ```
 
+### Pi
+
+Recommended full Pi package install:
+
+```bash
+./scripts/install-pi-package.sh --global
+# or, for project-local Pi package install
+./scripts/install-pi-package.sh --local
+```
+
+Manual single-skill Pi install from the package mirror:
+
+```bash
+./scripts/sync-pi-package-skills.sh
+mkdir -p .pi/skills/create-plan
+cp -R pi-package/skills/create-plan/* .pi/skills/create-plan/
+mkdir -p .pi/skills/reviewer-runtime/pi
+cp -R skills/reviewer-runtime/pi/* .pi/skills/reviewer-runtime/pi/
+chmod +x .pi/skills/reviewer-runtime/pi/*.sh
+```
+
+Global manual installs use `~/.pi/agent/skills/create-plan/` and `~/.pi/agent/skills/reviewer-runtime/pi/` instead of `.pi/skills/...`.
+
+Pi workflow skills also require Superpowers. See [PI-SUPERPOWERS.md](./PI-SUPERPOWERS.md) and [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md).
+
 ## Verify Installation
 
 ```bash
@@ -94,10 +121,12 @@ test -f ~/.codex/skills/create-plan/SKILL.md || true
 test -f ~/.claude/skills/create-plan/SKILL.md || true
 test -f ~/.config/opencode/skills/create-plan/SKILL.md || true
 test -f .cursor/skills/create-plan/SKILL.md || test -f ~/.cursor/skills/create-plan/SKILL.md || true
+test -f .pi/skills/create-plan/SKILL.md || test -f ~/.pi/agent/skills/create-plan/SKILL.md || true
 test -x ~/.codex/skills/reviewer-runtime/run-review.sh || true
 test -x ~/.claude/skills/reviewer-runtime/run-review.sh || true
 test -x ~/.config/opencode/skills/reviewer-runtime/run-review.sh || true
 test -x .cursor/skills/reviewer-runtime/run-review.sh || test -x ~/.cursor/skills/reviewer-runtime/run-review.sh || true
+test -x .pi/skills/reviewer-runtime/pi/run-review.sh || test -x ~/.pi/agent/skills/reviewer-runtime/pi/run-review.sh || true
 ```
 
 Verify Superpowers dependencies exist in your agent skills root:
@@ -110,6 +139,8 @@ Verify Superpowers dependencies exist in your agent skills root:
 - OpenCode: `~/.config/opencode/skills/superpowers/writing-plans/SKILL.md`
 - Cursor: `.cursor/skills/superpowers/skills/brainstorming/SKILL.md` or `~/.cursor/skills/superpowers/skills/brainstorming/SKILL.md`
 - Cursor: `.cursor/skills/superpowers/skills/writing-plans/SKILL.md` or `~/.cursor/skills/superpowers/skills/writing-plans/SKILL.md`
+- Pi: `.pi/skills/superpowers/brainstorming/SKILL.md` or `~/.pi/agent/skills/superpowers/brainstorming/SKILL.md` or `~/.agents/skills/superpowers/brainstorming/SKILL.md`
+- Pi: `.pi/skills/superpowers/writing-plans/SKILL.md` or `~/.pi/agent/skills/superpowers/writing-plans/SKILL.md` or `~/.agents/skills/superpowers/writing-plans/SKILL.md`
 
 ## Key Behavior
 
@@ -134,7 +165,7 @@ Verify Superpowers dependencies exist in your agent skills root:
 
 After the plan is created (design + milestones + stories), the skill sends it to a second model for review:
 
-1. **Configure** — user picks a reviewer CLI (`codex`, `claude`, `cursor`), a model, and optional max rounds (default 10), or skips
+1. **Configure** — user picks a reviewer CLI (`codex`, `claude`, `cursor`, `pi`), a model, and optional max rounds (default 10), or skips
 2. **Prepare** — plan payload and a bash reviewer command script are written to temp files
 3. **Run** — the command script is executed through `reviewer-runtime/run-review.sh` when installed
 4. **Feedback** — reviewer evaluates correctness, risks, missing steps, alternatives, security, and returns `## Summary`, `## Findings`, and `## Verdict`
@@ -188,13 +219,24 @@ ts=<ISO-8601> level=<info|warn|error> state=<running-silent|running-active|in-pr
 | `codex` | `codex exec -m <model> -s read-only` | Yes (`codex exec resume <id>`) | `-s read-only` |
 | `claude` | `claude -p --model <model> --strict-mcp-config --setting-sources user` | No (fresh call each round) | `--strict-mcp-config --setting-sources user` |
 | `cursor` | `cursor-agent -p --mode=ask --model <model> --trust --output-format json` | Yes (`--resume <id>`) | `--mode=ask` |
+| `pi` | See [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md) | No (fresh call each round) | `--tools read,grep,find,ls` |
 
-For all three CLIs, the preferred execution path is:
+For all supported reviewer CLIs, the preferred execution path is:
 
 1. write the reviewer command to a bash script
 2. run that script through `reviewer-runtime/run-review.sh`
 3. fall back to direct synchronous execution only if the helper is missing or not executable
 
+
+
+## Pi Reviewer Support
+
+All workflow variants can use Pi itself as a reviewer CLI. Use `pi/<pi-model-name>` shorthand, for example `pi/claude-opus-4-7`; this means `REVIEWER_CLI=pi` and `REVIEWER_MODEL=claude-opus-4-7`. Provider-qualified or multi-slash Pi model IDs are preserved after the first `pi/` prefix, for example `pi/anthropic/claude-opus-4-7`.
+
+The canonical isolated read-only Pi reviewer flag contract lives in [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md). This workflow passes the plan review payload at `/tmp/plan-${REVIEW_ID}.md` and expects the standard `## Summary`, `## Findings`, and `## Verdict` response. Pi reviewer output is captured as markdown stdout, not JSON.
+
+If the Pi reviewer model or provider is unavailable, surface the helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ## Notifications
 
 - Telegram is the only supported notification path.

docs/CURSOR.md

@@ -0,0 +1,103 @@
+# Cursor Manual Install
+
+## Skill Roots
+
+Cursor supports repo-local and global skills:
+
+```bash
+.cursor/skills/<skill-name>/
+~/.cursor/skills/<skill-name>/
+```
+
+Manual repo-local install example:
+
+```bash
+mkdir -p .cursor/skills/create-plan
+cp -R skills/create-plan/cursor/* .cursor/skills/create-plan/
+```
+
+Global install example:
+
+```bash
+mkdir -p ~/.cursor/skills/create-plan
+cp -R skills/create-plan/cursor/* ~/.cursor/skills/create-plan/
+```
+
+Cursor variants exist for `atlassian`, `create-plan`, `do-task`, `implement-plan`, and `web-automation`.
+
+Web automation repo-local install:
+
+```bash
+mkdir -p .cursor/skills/web-automation
+cp -R skills/web-automation/cursor/* .cursor/skills/web-automation/
+cd .cursor/skills/web-automation/scripts
+pnpm install
+npx cloakbrowser install
+pnpm approve-builds
+pnpm rebuild better-sqlite3 esbuild
+```
+
+Global web automation installs use `~/.cursor/skills/web-automation/` instead.
+
+## Reviewer Runtime
+
+Repo-local:
+
+```bash
+mkdir -p .cursor/skills/reviewer-runtime
+cp skills/reviewer-runtime/run-review.sh .cursor/skills/reviewer-runtime/
+cp skills/reviewer-runtime/notify-telegram.sh .cursor/skills/reviewer-runtime/
+chmod +x .cursor/skills/reviewer-runtime/*.sh
+```
+
+Global uses `~/.cursor/skills/reviewer-runtime/` instead.
+
+## Superpowers
+
+Cursor can discover Superpowers from the Cursor plugin cache or from manual
+repo-local/global skill roots. Prefer the plugin install when it is present;
+do not also install a manual Superpowers copy, or Cursor may show each
+Superpowers skill twice.
+
+```bash
+.cursor/plugins/cache/cursor-public/superpowers/<revision>/skills/<superpower>/SKILL.md
+~/.cursor/plugins/cache/cursor-public/superpowers/<revision>/skills/<superpower>/SKILL.md
+.cursor/skills/superpowers/skills/<superpower>/SKILL.md
+~/.cursor/skills/superpowers/skills/<superpower>/SKILL.md
+```
+
+Manual symlink, only when the Cursor plugin is not installed:
+
+```bash
+mkdir -p .cursor/skills/superpowers
+ln -s /absolute/path/to/obra/superpowers/skills .cursor/skills/superpowers/skills
+```
+
+## Cursor Reviewer Notes
+
+Cursor reviewer calls use JSON output and require `jq` where the skill variant parses reviewer responses:
+
+```bash
+cursor-agent --version
+jq --version
+```
+
+Reviewer calls must use `--mode=ask --trust --output-format json`, never write-capable modes.
+
+## Verify
+
+Repo-local scope:
+
+```bash
+cursor-agent --version
+test -f .cursor/skills/create-plan/SKILL.md
+test -f .cursor/skills/superpowers/skills/brainstorming/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/brainstorming/SKILL.md' -print -quit 2>/dev/null | grep -q .
+```
+
+Global scope:
+
+```bash
+cursor-agent --version
+test -f ~/.cursor/skills/create-plan/SKILL.md
+test -f ~/.cursor/skills/superpowers/skills/brainstorming/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/brainstorming/SKILL.md' -print -quit 2>/dev/null | grep -q .
+```

docs/DO-TASK.md

@@ -25,6 +25,7 @@ Execute a single user-supplied prompt end-to-end with **two reviewer loops** (pl
   - Claude Code: `~/.claude/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
   - OpenCode: `~/.config/opencode/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
   - Cursor: `.cursor/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}` (repo-local, preferred) or `~/.cursor/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}` (global fallback)
+  - Pi: `.pi/skills/reviewer-runtime/pi/{run-review.sh,notify-telegram.sh}` (repo-local) or `~/.pi/agent/skills/reviewer-runtime/pi/{run-review.sh,notify-telegram.sh}` (global)
 - Variant-specific prerequisites:
   - **Claude Code:** `claude --version`, explicit `Skill`-tool invocation of sub-skills.
   - **Codex:** `codex --version`; `~/.agents/skills/superpowers -> ~/.codex/superpowers/skills` symlink present.
@@ -38,6 +39,7 @@ Dependency-missing messages are variant-specific:
 - **Codex:** `Missing dependency: [specific missing item]. Install required Superpowers skills (https://github.com/obra/superpowers) and the reviewer-runtime helper, then retry.`
 - **Cursor:** `Missing dependency: [specific missing item]. Install Cursor Agent CLI, jq, and Superpowers skills under .cursor/skills/ or ~/.cursor/skills/, then retry.`
 - **OpenCode:** `Missing dependency: [specific missing item]. Install required OpenCode Superpowers skills (https://github.com/obra/superpowers, OpenCode setup) and the reviewer-runtime helper, then retry.`
+- **Pi:** `Missing dependency: [specific missing item]. Install Pi, required Superpowers skills, and the Pi reviewer-runtime helper, then retry.`
 
 ### Reviewer CLI Requirements
 
@@ -49,6 +51,7 @@ One of these CLIs must be installed to drive either of the two review loops:
 | `claude` | `npm install -g @anthropic-ai/claude-code` | `claude --version` | `--strict-mcp-config --setting-sources user` | No (fresh call each round) |
 | `cursor` | `curl https://cursor.com/install -fsS \| bash` | `cursor-agent --version` (binary: `cursor-agent`; alias `cursor agent` also works) | `--mode=ask` | Yes (`--resume <id>`) |
 | `opencode` | `brew install opencode` or your package manager | `opencode --version` | `--agent plan` | Opt-in (`-s <id>`; fresh call is the default) |
+| `pi` | Install Pi coding agent | `pi --version`; list models with `pi --list-models [search]` | `--tools read,grep,find,ls` | No (fresh call each round) |
 
 The reviewer CLI is independent of which agent is running the skill — e.g., Claude Code can send both the plan and the implementation to Codex for review.
 
@@ -103,6 +106,31 @@ mkdir -p ~/.cursor/skills/reviewer-runtime
 cp -R skills/reviewer-runtime/* ~/.cursor/skills/reviewer-runtime/
 ```
 
+### Pi
+
+Recommended full Pi package install:
+
+```bash
+./scripts/install-pi-package.sh --global
+# or, for project-local Pi package install
+./scripts/install-pi-package.sh --local
+```
+
+Manual single-skill Pi install from the package mirror:
+
+```bash
+./scripts/sync-pi-package-skills.sh
+mkdir -p .pi/skills/do-task
+cp -R pi-package/skills/do-task/* .pi/skills/do-task/
+mkdir -p .pi/skills/reviewer-runtime/pi
+cp -R skills/reviewer-runtime/pi/* .pi/skills/reviewer-runtime/pi/
+chmod +x .pi/skills/reviewer-runtime/pi/*.sh
+```
+
+Global manual installs use `~/.pi/agent/skills/do-task/` and `~/.pi/agent/skills/reviewer-runtime/pi/` instead of `.pi/skills/...`.
+
+Pi workflow skills also require Superpowers. See [PI-SUPERPOWERS.md](./PI-SUPERPOWERS.md) and [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md).
+
 ## Verify Installation
 
 Run the per-variant checks for everything the corresponding `SKILL.md` enforces. Each check is structured: (1) CLI binary version, (2) skill file presence, (3) reviewer-runtime + notifier helper presence, (4) Superpowers sub-skill discovery, (5) variant-specific extras.
@@ -162,6 +190,19 @@ test -f .cursor/skills/superpowers/skills/verification-before-completion/SKILL.m
 test -f .cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md
 ```
 
+### Pi
+
+```bash
+pi --version
+test -f .pi/skills/do-task/SKILL.md || test -f ~/.pi/agent/skills/do-task/SKILL.md
+test -x .pi/skills/reviewer-runtime/pi/run-review.sh || test -x ~/.pi/agent/skills/reviewer-runtime/pi/run-review.sh
+test -x .pi/skills/reviewer-runtime/pi/notify-telegram.sh || test -x ~/.pi/agent/skills/reviewer-runtime/pi/notify-telegram.sh
+test -f .pi/skills/superpowers/brainstorming/SKILL.md || test -f ~/.pi/agent/skills/superpowers/brainstorming/SKILL.md || test -f ~/.agents/skills/superpowers/brainstorming/SKILL.md
+test -f .pi/skills/superpowers/test-driven-development/SKILL.md || test -f ~/.pi/agent/skills/superpowers/test-driven-development/SKILL.md || test -f ~/.agents/skills/superpowers/test-driven-development/SKILL.md
+test -f .pi/skills/superpowers/verification-before-completion/SKILL.md || test -f ~/.pi/agent/skills/superpowers/verification-before-completion/SKILL.md || test -f ~/.agents/skills/superpowers/verification-before-completion/SKILL.md
+test -f .pi/skills/superpowers/finishing-a-development-branch/SKILL.md || test -f ~/.pi/agent/skills/superpowers/finishing-a-development-branch/SKILL.md || test -f ~/.agents/skills/superpowers/finishing-a-development-branch/SKILL.md
+```
+
 ## Key Behavior
 
 - Creates one persistent plan artifact at `ai_plan/YYYY-MM-DD-<slug>/task-plan.md`.
@@ -290,6 +331,7 @@ The user answers `yes` / `no` / `redact`:
 | `claude` | `claude -p "<prompt>" --model <model> --strict-mcp-config --setting-sources user` | Fresh call with prior-round context summary | `cp <runner.out> <out.md>` |
 | `cursor` | `cursor-agent -p --mode=ask --model <model> --trust --output-format json "<prompt>" > <out.json>` | `cursor-agent --resume <id> -p --mode=ask --model <model> --trust --output-format json "<prompt>" > <out.json>` | `jq -r '.result' <out.json> > <out.md>` |
 | `opencode` | `opencode run -m <provider>/<model> --agent plan --format json "<prompt>" > <out.json>` | Fresh call (default) OR `opencode run -s <id> -m <provider>/<model> --agent plan --format json "<prompt>" > <out.json>` (opt-in) | `jq -r '.[] \| select(.type == "message" and .role == "assistant") \| .content' <out.json> > <out.md>` |
+| `pi` | See [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md) | Fresh call | Markdown stdout copied to `<out.md>` |
 
 For all four CLIs, the preferred execution path is:
 
@@ -297,6 +339,16 @@ For all four CLIs, the preferred execution path is:
 2. Run that script through `reviewer-runtime/run-review.sh`.
 3. Fall back to direct synchronous execution only if the helper is missing or not executable.
 
+
+
+## Pi Reviewer Support
+
+All workflow variants can use Pi itself as a reviewer CLI. Use `pi/<pi-model-name>` shorthand, for example `pi/claude-opus-4-7`; this means `REVIEWER_CLI=pi` and `REVIEWER_MODEL=claude-opus-4-7`. Provider-qualified or multi-slash Pi model IDs are preserved after the first `pi/` prefix, for example `pi/anthropic/claude-opus-4-7`.
+
+The canonical isolated read-only Pi reviewer flag contract lives in [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md). This workflow passes the plan and implementation review payload at `/tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md` and expects the standard `## Summary`, `## Findings`, and `## Verdict` response. Pi reviewer output is captured as markdown stdout, not JSON.
+
+If the Pi reviewer model or provider is unavailable, surface the helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ## Notifications
 
 - Telegram is the only supported notification path.

docs/IMPLEMENT-PLAN.md

@@ -21,10 +21,11 @@ Execute an existing plan (created by `create-plan`) in an isolated git worktree,
   - `~/.agents/skills/superpowers -> ~/.codex/superpowers/skills`
 - For Cursor, skills must be installed under `.cursor/skills/` (repo-local) or `~/.cursor/skills/` (global)
 - Shared reviewer runtime must be installed beside agent skills when using reviewer CLIs:
-  - Codex: `~/.codex/skills/reviewer-runtime/run-review.sh`
-  - Claude Code: `~/.claude/skills/reviewer-runtime/run-review.sh`
-  - OpenCode: `~/.config/opencode/skills/reviewer-runtime/run-review.sh`
-  - Cursor: `.cursor/skills/reviewer-runtime/run-review.sh` or `~/.cursor/skills/reviewer-runtime/run-review.sh`
+  - Codex: `~/.codex/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
+  - Claude Code: `~/.claude/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
+  - OpenCode: `~/.config/opencode/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
+  - Cursor: `.cursor/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}` or `~/.cursor/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`
+  - Pi: `.pi/skills/reviewer-runtime/pi/{run-review.sh,notify-telegram.sh}` or `~/.pi/agent/skills/reviewer-runtime/pi/{run-review.sh,notify-telegram.sh}`
 - Telegram notification setup is documented in [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md)
 
 If dependencies are missing, stop and return:
@@ -40,6 +41,7 @@ To use the iterative milestone review feature, one of these CLIs must be install
 | `codex` | `npm install -g @openai/codex` | `codex --version` |
 | `claude` | `npm install -g @anthropic-ai/claude-code` | `claude --version` |
 | `cursor` | `curl https://cursor.com/install -fsS \| bash` | `cursor-agent --version` (binary: `cursor-agent`; alias `cursor agent` also works) |
+| `pi` | Install Pi coding agent | `pi --version`; list models with `pi --list-models [search]` |
 
 The reviewer CLI is independent of which agent is running the implementation — e.g., Claude Code can send milestones to Codex for review, and vice versa.
 
@@ -94,6 +96,31 @@ mkdir -p ~/.cursor/skills/reviewer-runtime
 cp -R skills/reviewer-runtime/* ~/.cursor/skills/reviewer-runtime/
 ```
 
+### Pi
+
+Recommended full Pi package install:
+
+```bash
+./scripts/install-pi-package.sh --global
+# or, for project-local Pi package install
+./scripts/install-pi-package.sh --local
+```
+
+Manual single-skill Pi install from the package mirror:
+
+```bash
+./scripts/sync-pi-package-skills.sh
+mkdir -p .pi/skills/implement-plan
+cp -R pi-package/skills/implement-plan/* .pi/skills/implement-plan/
+mkdir -p .pi/skills/reviewer-runtime/pi
+cp -R skills/reviewer-runtime/pi/* .pi/skills/reviewer-runtime/pi/
+chmod +x .pi/skills/reviewer-runtime/pi/*.sh
+```
+
+Global manual installs use `~/.pi/agent/skills/implement-plan/` and `~/.pi/agent/skills/reviewer-runtime/pi/` instead of `.pi/skills/...`.
+
+Pi workflow skills also require Superpowers. See [PI-SUPERPOWERS.md](./PI-SUPERPOWERS.md) and [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md).
+
 ## Verify Installation
 
 ```bash
@@ -101,10 +128,12 @@ test -f ~/.codex/skills/implement-plan/SKILL.md || true
 test -f ~/.claude/skills/implement-plan/SKILL.md || true
 test -f ~/.config/opencode/skills/implement-plan/SKILL.md || true
 test -f .cursor/skills/implement-plan/SKILL.md || test -f ~/.cursor/skills/implement-plan/SKILL.md || true
+test -f .pi/skills/implement-plan/SKILL.md || test -f ~/.pi/agent/skills/implement-plan/SKILL.md || true
 test -x ~/.codex/skills/reviewer-runtime/run-review.sh || true
 test -x ~/.claude/skills/reviewer-runtime/run-review.sh || true
 test -x ~/.config/opencode/skills/reviewer-runtime/run-review.sh || true
 test -x .cursor/skills/reviewer-runtime/run-review.sh || test -x ~/.cursor/skills/reviewer-runtime/run-review.sh || true
+test -x .pi/skills/reviewer-runtime/pi/run-review.sh || test -x ~/.pi/agent/skills/reviewer-runtime/pi/run-review.sh || true
 ```
 
 Verify Superpowers execution dependencies exist in your agent skills root:
@@ -125,6 +154,10 @@ Verify Superpowers execution dependencies exist in your agent skills root:
 - Cursor: `.cursor/skills/superpowers/skills/using-git-worktrees/SKILL.md` or `~/.cursor/skills/superpowers/skills/using-git-worktrees/SKILL.md`
 - Cursor: `.cursor/skills/superpowers/skills/verification-before-completion/SKILL.md` or `~/.cursor/skills/superpowers/skills/verification-before-completion/SKILL.md`
 - Cursor: `.cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md` or `~/.cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md`
+- Pi: `.pi/skills/superpowers/executing-plans/SKILL.md` or `~/.pi/agent/skills/superpowers/executing-plans/SKILL.md` or `~/.agents/skills/superpowers/executing-plans/SKILL.md`
+- Pi: `.pi/skills/superpowers/using-git-worktrees/SKILL.md` or `~/.pi/agent/skills/superpowers/using-git-worktrees/SKILL.md` or `~/.agents/skills/superpowers/using-git-worktrees/SKILL.md`
+- Pi: `.pi/skills/superpowers/verification-before-completion/SKILL.md` or `~/.pi/agent/skills/superpowers/verification-before-completion/SKILL.md` or `~/.agents/skills/superpowers/verification-before-completion/SKILL.md`
+- Pi: `.pi/skills/superpowers/finishing-a-development-branch/SKILL.md` or `~/.pi/agent/skills/superpowers/finishing-a-development-branch/SKILL.md` or `~/.agents/skills/superpowers/finishing-a-development-branch/SKILL.md`
 
 ## Key Behavior
 
@@ -146,14 +179,14 @@ Verify Superpowers execution dependencies exist in your agent skills root:
 
 After each milestone is implemented and verified, the skill sends it to a second model for review:
 
-1. **Configure** — user picks a reviewer CLI (`codex`, `claude`, `cursor`) and model, or skips
+1. **Configure** — user picks a reviewer CLI (`codex`, `claude`, `cursor`, `pi`) and model, or skips
 2. **Prepare** — milestone payload and a bash reviewer command script are written to temp files
 3. **Run** — the command script is executed through `reviewer-runtime/run-review.sh` when installed
 4. **Feedback** — reviewer evaluates correctness, acceptance criteria, code quality, test coverage, security, and returns `## Summary`, `## Findings`, and `## Verdict`
 5. **Prioritize** — findings are ordered `P0`, `P1`, `P2`, `P3`
 6. **Revise** — the implementing agent addresses findings in priority order, re-verifies, and re-submits
 7. **Repeat** — up to max rounds (default 10) until the reviewer returns `VERDICT: APPROVED`
-7. **Approve** — milestone is marked approved in `story-tracker.md`
+8. **Approve** — milestone is marked approved in `story-tracker.md`
 
 ### Reviewer Output Contract
 
@@ -200,13 +233,24 @@ ts=<ISO-8601> level=<info|warn|error> state=<running-silent|running-active|in-pr
 | `codex` | `codex exec -m <model> -s read-only` | Yes (`codex exec resume <id>`) | `-s read-only` |
 | `claude` | `claude -p --model <model> --strict-mcp-config --setting-sources user` | No (fresh call each round) | `--strict-mcp-config --setting-sources user` |
 | `cursor` | `cursor-agent -p --mode=ask --model <model> --trust --output-format json` | Yes (`--resume <id>`) | `--mode=ask` |
+| `pi` | See [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md) | No (fresh call each round) | `--tools read,grep,find,ls` |
 
-For all three CLIs, the preferred execution path is:
+For all supported reviewer CLIs, the preferred execution path is:
 
 1. write the reviewer command to a bash script
 2. run that script through `reviewer-runtime/run-review.sh`
 3. fall back to direct synchronous execution only if the helper is missing or not executable
 
+
+
+## Pi Reviewer Support
+
+All workflow variants can use Pi itself as a reviewer CLI. Use `pi/<pi-model-name>` shorthand, for example `pi/claude-opus-4-7`; this means `REVIEWER_CLI=pi` and `REVIEWER_MODEL=claude-opus-4-7`. Provider-qualified or multi-slash Pi model IDs are preserved after the first `pi/` prefix, for example `pi/anthropic/claude-opus-4-7`.
+
+The canonical isolated read-only Pi reviewer flag contract lives in [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md). This workflow passes the milestone review payload at `/tmp/milestone-${REVIEW_ID}.md` and expects the standard `## Summary`, `## Findings`, and `## Verdict` response. Pi reviewer output is captured as markdown stdout, not JSON.
+
+If the Pi reviewer model or provider is unavailable, surface the helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ## Notifications
 
 - Telegram is the only supported notification path.

docs/INSTALLER.md

@@ -0,0 +1,145 @@
+# Skill Manager Installer
+
+Use the skill manager wizard to install, update/reinstall, or remove skills from this repository for supported local coding clients.
+
+## Quick Start
+
+```bash
+./scripts/manage-skills.sh
+# or
+node scripts/manage-skills.mjs
+```
+
+Preview without changing files:
+
+```bash
+node scripts/manage-skills.mjs --dry-run
+```
+
+Drive the planner non-interactively:
+
+```bash
+node scripts/manage-skills.mjs --plan-only --answers answers.json
+```
+
+## Supported Clients
+
+The wizard detects these clients by CLI and known skill directories:
+
+| Client | CLI check | Default skill root |
+|---|---|---|
+| Codex | `codex --version` | `~/.codex/skills` |
+| Claude Code | `claude --version` | `~/.claude/skills` |
+| Cursor | `cursor-agent --version` then `cursor agent --version` | `.cursor/skills` or `~/.cursor/skills` |
+| OpenCode | `opencode --version` | `~/.config/opencode/skills` |
+| Pi | `pi --version` | package mode or `.pi/skills` / `~/.pi/agent/skills` |
+
+## Actions
+
+For each selected client/scope, the wizard lists repository skills and offers actions based on installed state:
+
+- missing skill: `install` or `skip`
+- installed/stale/unknown skill: `update`, `reinstall`, `remove`, or `skip`
+- unsupported variant: reported as `unsupported` and skipped
+
+The current skill matrix is:
+
+| Skill | Codex | Claude Code | Cursor | OpenCode | Pi |
+|---|---:|---:|---:|---:|---:|
+| `atlassian` | yes | yes | yes | yes | yes |
+| `create-plan` | yes | yes | yes | yes | yes |
+| `do-task` | yes | yes | yes | yes | yes |
+| `implement-plan` | yes | yes | yes | yes | yes |
+| `web-automation` | yes | yes | yes | yes | yes |
+
+## Superpowers Handling
+
+Workflow skills require Obra Superpowers:
+
+- `create-plan`: `brainstorming`, `writing-plans`
+- `implement-plan`: `executing-plans`, `using-git-worktrees`, `verification-before-completion`, `finishing-a-development-branch`
+- `do-task`: `brainstorming`, `test-driven-development`, `verification-before-completion`, `finishing-a-development-branch` plus conditional `using-git-worktrees`
+
+When a selected install/update needs Superpowers and none are detected for the target client/scope, the wizard asks whether to install them. It asks for an absolute path to an Obra Superpowers `skills` directory and lets you choose symlink or copy:
+
+- symlink is recommended because updates to the source checkout are immediately visible
+- copy is more self-contained but must be updated manually
+
+For Cursor, the wizard installs Superpowers at `<scope>/superpowers/skills` because Cursor variants expect `superpowers/skills/<skill>/SKILL.md`. Codex commonly reuses the shared `~/.agents/skills/superpowers` convention documented in [CODEX.md](./CODEX.md).
+
+When removing the last repository workflow skill from a client/scope, the wizard asks whether to remove Superpowers for that variant too.
+
+## Reviewer Runtime Helpers
+
+Workflow skills install/update the reviewer-runtime helper bundle automatically when needed:
+
+- non-Pi clients receive `run-review.sh` and `notify-telegram.sh` from `skills/reviewer-runtime/`
+- Pi receives `run-review.sh` and `notify-telegram.sh` from `skills/reviewer-runtime/pi/`
+- diagnostics tests and nested Pi helper directories are not copied into non-Pi installs
+
+## Pi Package Mode
+
+Pi can be managed as a package install or by manual skill copy. Package mode always manages the full Pi bundle; per-skill prompts and `--skill` narrowing are ignored for `packageGlobal` and `packageLocal`.
+
+Package-mode actions:
+
+- `install`: register the package if needed, list bundled skills, and skip already-bootstrapped runtimes.
+- `update`: sync the Pi package mirror, reinstall the package registration, and rerun runtime dependency bootstrapping.
+- `reinstall`: same behavior as `update`, kept for action parity with manual skill scopes.
+- `remove`: unregister the package with `pi remove`; this does not delete repo files or `node_modules`.
+
+Examples:
+
+```bash
+node scripts/manage-skills.mjs --client pi --scope packageGlobal --pi-package --action install --yes
+node scripts/manage-skills.mjs --client pi --scope packageLocal --pi-package --action install --yes
+node scripts/manage-skills.mjs --client pi --scope packageGlobal --pi-package --action update --yes
+node scripts/manage-skills.mjs --client pi --scope packageGlobal --pi-package --action remove --yes
+```
+
+The compatibility script remains available:
+
+```bash
+./scripts/install-pi-package.sh --global
+./scripts/install-pi-package.sh --local
+```
+
+## Answers JSON
+
+`--plan-only --answers` accepts this shape:
+
+```json
+{
+  "selections": [
+    {
+      "clientId": "codex",
+      "scope": "global",
+      "actions": {
+        "create-plan": "install",
+        "web-automation": "skip"
+      }
+    }
+  ]
+}
+```
+
+The output is JSON containing planned operations, dependency prompts, and final-report rows. No filesystem changes are made.
+
+## Final Report
+
+The final report uses these columns:
+
+| Column | Meaning |
+|---|---|
+| client | client id |
+| scope | selected target scope |
+| skill/helper | skill name or helper bundle |
+| action | install, update, reinstall, remove, sync, bootstrap, etc. |
+| status | `ok`, `skipped`, `failed`, or `warning` |
+| details | target path or error detail |
+
+Exit code is non-zero if any selected operation fails.
+
+Dangling symlink warnings are surfaced as `warning` rows. For example, if a previously symlinked Superpowers source has moved or been deleted, the final report keeps the operation non-destructive and shows the dangling symlink target in `details` so you can repair or remove it deliberately.
+
+See [PI.md](./PI.md) for Pi package layout details and mirror maintenance.

docs/OPENCODE.md

@@ -0,0 +1,72 @@
+# OpenCode Manual Install
+
+## Skill Root
+
+OpenCode skills are installed under:
+
+```bash
+~/.config/opencode/skills/<skill-name>/
+```
+
+Manual install example:
+
+```bash
+mkdir -p ~/.config/opencode/skills/implement-plan
+cp -R skills/implement-plan/opencode/* ~/.config/opencode/skills/implement-plan/
+```
+
+Use `skills/<skill>/opencode/*` for each supported skill.
+
+## Reviewer Runtime
+
+```bash
+mkdir -p ~/.config/opencode/skills/reviewer-runtime
+cp skills/reviewer-runtime/run-review.sh ~/.config/opencode/skills/reviewer-runtime/
+cp skills/reviewer-runtime/notify-telegram.sh ~/.config/opencode/skills/reviewer-runtime/
+chmod +x ~/.config/opencode/skills/reviewer-runtime/*.sh
+```
+
+## Superpowers
+
+OpenCode can discover Superpowers from the shared agents skill root or the
+OpenCode-specific skills root:
+
+```bash
+~/.agents/skills/superpowers
+~/.config/opencode/skills/superpowers
+```
+
+OpenCode's native setup commonly exposes the shared agents root:
+
+```bash
+ln -s /absolute/path/to/obra/superpowers/skills ~/.agents/skills/superpowers
+```
+
+OpenCode-specific setup is also supported:
+
+```bash
+ln -s /absolute/path/to/obra/superpowers/skills ~/.config/opencode/skills/superpowers
+```
+
+Verify:
+
+```bash
+test -f ~/.agents/skills/superpowers/brainstorming/SKILL.md || test -f ~/.config/opencode/skills/superpowers/brainstorming/SKILL.md
+test -f ~/.agents/skills/superpowers/verification-before-completion/SKILL.md || test -f ~/.config/opencode/skills/superpowers/verification-before-completion/SKILL.md
+```
+
+## OpenCode Reviewer Notes
+
+OpenCode reviewer calls use the built-in read-oriented plan agent:
+
+```bash
+opencode run -m <provider>/<model> --agent plan --format json "review prompt"
+```
+
+## Verify
+
+```bash
+opencode --version
+test -f ~/.config/opencode/skills/implement-plan/SKILL.md
+test -x ~/.config/opencode/skills/reviewer-runtime/run-review.sh
+```

docs/PI-COMMON-REVIEWER.md

@@ -16,6 +16,18 @@ The canonical isolated, read-only reviewer command is:
 pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files --model "$REVIEWER_MODEL" --tools read,grep,find,ls -p "Read the review payload and return the requested verdict."
 ```
 
+Workflow docs should link to this section instead of restating the full flag contract. Each workflow substitutes its own payload path and verdict prompt:
+
+- `create-plan`: `/tmp/plan-${REVIEW_ID}.md`
+- `implement-plan`: `/tmp/milestone-${REVIEW_ID}.md`
+- `do-task`: `/tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md`
+
+User-facing shorthand `pi/<pi-model-name>` means `REVIEWER_CLI=pi` and `REVIEWER_MODEL=<pi-model-name>`. Split only on the first slash when the prefix before that slash is exactly `pi`; keep the remainder verbatim. Examples:
+
+- `pi/claude-opus-4-7` -> `claude-opus-4-7`
+- `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`
+- `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`
+
 Pi reviewer calls must stay isolated from the main workflow:
 
 - Use `--no-session` so the reviewer does not continue or persist the workflow session.

docs/PI.md

@@ -52,7 +52,16 @@ Workflow-heavy Pi skills split their shared setup across two docs:
 
 ## Package Install
 
-The user-facing install flow is the repo-owned installer script, not a raw `pi install` command.
+The multi-client skill manager can guide Pi install/update/remove operations alongside the other supported clients:
+
+```bash
+./scripts/manage-skills.sh
+node scripts/manage-skills.mjs --dry-run
+```
+
+The compatibility Pi package installer remains available for the focused Pi package path.
+
+The user-facing Pi package install flow is the repo-owned installer script, not a raw `pi install` command.
 
 Global install from a cloned checkout:
 
@@ -115,7 +124,7 @@ When a source Pi variant changes:
 npm pack --dry-run --json
 ```
 
-The installer intentionally does not run sync. It assumes the checked-in `pi-package/skills/*` mirror is already current.
+The focused `scripts/install-pi-package.sh` installer intentionally does not run sync. It assumes the checked-in `pi-package/skills/*` mirror is already current. The multi-client skill manager plans a sync step before manual Pi skill-copy operations so manual installs use the current package-facing mirror.
 
 The verifier is responsible for catching:
 

docs/README.md

@@ -2,8 +2,20 @@
 
 This directory contains user-facing docs for each skill.
 
+## Recommended Reading Flow
+
+1. Use [INSTALLER.md](./INSTALLER.md) for the guided install/update/remove wizard (`./scripts/manage-skills.sh`).
+2. Use the client docs ([CODEX.md](./CODEX.md), [CLAUDE-CODE.md](./CLAUDE-CODE.md), [CURSOR.md](./CURSOR.md), [OPENCODE.md](./OPENCODE.md)) for manual installs, and [PI.md](./PI.md) for Pi overview, package install, and manual single-skill copy.
+3. Use the skill docs below for skill-specific requirements, runtime setup, verification, and examples.
+4. For Pi workflow skills, also consult [PI-SUPERPOWERS.md](./PI-SUPERPOWERS.md) and [PI-COMMON-REVIEWER.md](./PI-COMMON-REVIEWER.md).
+
 ## Index
 
+- [INSTALLER.md](./INSTALLER.md) — Skill manager wizard for installing, updating/reinstalling, and removing repo skills across supported clients.
+- [CODEX.md](./CODEX.md) — Manual install instructions for Codex variants.
+- [CLAUDE-CODE.md](./CLAUDE-CODE.md) — Manual install instructions for Claude Code variants.
+- [CURSOR.md](./CURSOR.md) — Manual install instructions for Cursor variants.
+- [OPENCODE.md](./OPENCODE.md) — Manual install instructions for OpenCode variants.
 - [ATLASSIAN.md](./ATLASSIAN.md) — Includes requirements, generated bundle sync, install, auth, safety rules, and usage examples for the Atlassian skill.
 - [CREATE-PLAN.md](./CREATE-PLAN.md) — Includes requirements, install, verification, and execution workflow for create-plan.
 - [DO-TASK.md](./DO-TASK.md) — Single-prompt end-to-end execution with dual reviewer loops (plan + implementation), TDD-first, single task commit. Sibling of create-plan/implement-plan scoped to ad-hoc tasks.

docs/WEB-AUTOMATION.md

@@ -48,6 +48,22 @@ pnpm approve-builds
 pnpm rebuild better-sqlite3 esbuild
 ```
 
+### Cursor
+
+Repo-local install:
+
+```bash
+mkdir -p .cursor/skills/web-automation
+cp -R skills/web-automation/cursor/* .cursor/skills/web-automation/
+cd .cursor/skills/web-automation/scripts
+pnpm install
+npx cloakbrowser install
+pnpm approve-builds
+pnpm rebuild better-sqlite3 esbuild
+```
+
+Global installs use `~/.cursor/skills/web-automation/` instead.
+
 ### OpenCode
 
 ```bash
@@ -60,6 +76,31 @@ pnpm approve-builds
 pnpm rebuild better-sqlite3 esbuild
 ```
 
+### Pi
+
+Recommended full Pi package install:
+
+```bash
+./scripts/install-pi-package.sh --global
+# or, for project-local Pi package install
+./scripts/install-pi-package.sh --local
+```
+
+Manual single-skill Pi install from the package mirror:
+
+```bash
+./scripts/sync-pi-package-skills.sh
+mkdir -p .pi/skills/web-automation
+cp -R pi-package/skills/web-automation/* .pi/skills/web-automation/
+cd .pi/skills/web-automation/scripts
+pnpm install --frozen-lockfile
+npx cloakbrowser install
+pnpm approve-builds
+pnpm rebuild better-sqlite3 esbuild
+```
+
+Global manual installs use `~/.pi/agent/skills/web-automation/` instead of `.pi/skills/web-automation/`.
+
 ## Update To The Latest CloakBrowser
 
 Run inside the installed `scripts/` directory for the variant you are using:

package.json

@@ -12,9 +12,14 @@
   "files": [
     "README.md",
     "docs/ATLASSIAN.md",
+    "docs/CLAUDE-CODE.md",
+    "docs/CODEX.md",
     "docs/CREATE-PLAN.md",
+    "docs/CURSOR.md",
     "docs/DO-TASK.md",
     "docs/IMPLEMENT-PLAN.md",
+    "docs/INSTALLER.md",
+    "docs/OPENCODE.md",
     "docs/README.md",
     "docs/TELEGRAM-NOTIFICATIONS.md",
     "docs/PI.md",
@@ -25,9 +30,20 @@
     "pi-package/skills",
     "skills/reviewer-runtime/pi",
     "scripts/install-pi-package.sh",
+    "scripts/lib/skill-manager-core.mjs",
+    "scripts/manage-skills.mjs",
+    "scripts/manage-skills.sh",
     "scripts/sync-pi-package-skills.sh",
-    "scripts/verify-pi-resources.sh"
+    "scripts/verify-pi-resources.sh",
+    "scripts/verify-pi-workflows.sh",
+    "scripts/verify-reviewer-support.sh"
   ],
+  "scripts": {
+    "sync:pi": "./scripts/sync-pi-package-skills.sh",
+    "verify:pi": "./scripts/verify-pi-resources.sh && ./scripts/verify-pi-workflows.sh",
+    "verify:reviewers": "./scripts/verify-reviewer-support.sh",
+    "test:installer": "node --test scripts/tests/*.test.mjs"
+  },
   "pi": {
     "skills": [
       "./pi-package/skills/atlassian",

pi-package/skills/create-plan/SKILL.md

@@ -85,6 +85,14 @@ Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for the review loop.
 
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
 When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the pi model running this workflow. Use any configured pi model string, including provider-qualified model IDs. If the reviewer model or provider is unavailable, surface the review helper stderr/status and ask for a configured model; use `pi --list-models [search]` to inspect configured models.
 
 The pi reviewer command rendered into `/tmp/plan-review-${REVIEW_ID}.sh` must be isolated and read-only:

pi-package/skills/do-task/SKILL.md

@@ -103,6 +103,14 @@ Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS`.
 
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
 When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the pi model running this workflow. Use any configured pi model string, including provider-qualified model IDs. If the reviewer model or provider is unavailable, surface the review helper stderr/status and ask for a configured model; use `pi --list-models [search]` to inspect configured models.
 
 The pi reviewer command rendered into `/tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.sh` must be isolated and read-only:

pi-package/skills/implement-plan/SKILL.md

@@ -86,6 +86,14 @@ Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS`.
 
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
 When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the pi model running this workflow. Use any configured pi model string, including provider-qualified model IDs. If the reviewer model or provider is unavailable, surface the review helper stderr/status and ask for a configured model; use `pi --list-models [search]` to inspect configured models.
 
 The pi reviewer command rendered into `/tmp/milestone-review-${REVIEW_ID}.sh` must be isolated and read-only:

scripts/lib/skill-manager-core.mjs

@@ -0,0 +1,658 @@
+import { access, cp, lstat, mkdir, readFile, realpath, rm, stat, symlink, chmod, unlink, readdir } from "node:fs/promises";
+import { constants, existsSync } from "node:fs";
+import { homedir } from "node:os";
+import path from "node:path";
+import { spawnSync } from "node:child_process";
+
+export const REPO_ROOT = path.resolve(path.dirname(new URL(import.meta.url).pathname), "../..");
+
+export const CLIENTS = {
+  codex: {
+    id: "codex",
+    label: "Codex",
+    detectCommands: [["codex", "--version"]],
+    scopes: { global: { skillsRoot: "~/.codex/skills" } },
+    variant: "codex",
+    superpowers: {
+      roots: ["~/.agents/skills/superpowers", "~/.codex/superpowers/skills"],
+      layout: "flat",
+    },
+    reviewerRuntime: {
+      root: "~/.codex/skills/reviewer-runtime",
+      source: "skills/reviewer-runtime",
+      files: ["run-review.sh", "notify-telegram.sh"],
+    },
+  },
+  "claude-code": {
+    id: "claude-code",
+    label: "Claude Code",
+    detectCommands: [["claude", "--version"]],
+    scopes: { global: { skillsRoot: "~/.claude/skills" } },
+    variant: "claude-code",
+    superpowers: {
+      roots: ["~/.claude/skills/superpowers"],
+      layout: "flat",
+    },
+    reviewerRuntime: {
+      root: "~/.claude/skills/reviewer-runtime",
+      source: "skills/reviewer-runtime",
+      files: ["run-review.sh", "notify-telegram.sh"],
+    },
+  },
+  cursor: {
+    id: "cursor",
+    label: "Cursor",
+    detectCommands: [["cursor-agent", "--version"], ["cursor", "agent", "--version"]],
+    scopes: {
+      local: { skillsRoot: ".cursor/skills" },
+      global: { skillsRoot: "~/.cursor/skills" },
+    },
+    variant: "cursor",
+    superpowers: {
+      roots: [".cursor/skills/superpowers/skills", "~/.cursor/skills/superpowers/skills"],
+      layout: "cursor-nested",
+      targetSuffix: "superpowers/skills",
+    },
+    reviewerRuntime: {
+      root: "<skillsRoot>/reviewer-runtime",
+      source: "skills/reviewer-runtime",
+      files: ["run-review.sh", "notify-telegram.sh"],
+    },
+  },
+  opencode: {
+    id: "opencode",
+    label: "OpenCode",
+    detectCommands: [["opencode", "--version"]],
+    scopes: { global: { skillsRoot: "~/.config/opencode/skills" } },
+    variant: "opencode",
+    superpowers: {
+      roots: ["~/.agents/skills/superpowers", "~/.config/opencode/skills/superpowers"],
+      layout: "flat",
+    },
+    reviewerRuntime: {
+      root: "~/.config/opencode/skills/reviewer-runtime",
+      source: "skills/reviewer-runtime",
+      files: ["run-review.sh", "notify-telegram.sh"],
+    },
+  },
+  pi: {
+    id: "pi",
+    label: "Pi",
+    detectCommands: [["pi", "--version"]],
+    scopes: {
+      packageGlobal: { skillsRoot: "~/.pi/agent/skills", packageMode: true, piInstallArg: "" },
+      packageLocal: { skillsRoot: ".pi/skills", packageMode: true, piInstallArg: "-l" },
+      global: { skillsRoot: "~/.pi/agent/skills" },
+      local: { skillsRoot: ".pi/skills" },
+    },
+    variant: "pi",
+    superpowers: {
+      roots: ["~/.agents/skills/superpowers", "~/.pi/agent/skills/superpowers", ".pi/skills/superpowers"],
+      layout: "flat",
+    },
+    reviewerRuntime: {
+      root: "<skillsRoot>/reviewer-runtime/pi",
+      source: "skills/reviewer-runtime/pi",
+      files: ["run-review.sh", "notify-telegram.sh"],
+    },
+  },
+};
+
+export const SKILLS = {
+  atlassian: {
+    name: "atlassian",
+    variants: ["codex", "claude-code", "cursor", "opencode", "pi"],
+    requiresSuperpowers: [],
+    requiresReviewerRuntime: false,
+    bootstrap: "pnpm-install",
+    bootstrapPrerequisites: ["node>=20", "pnpm|corepack"],
+    auxiliary: { shared: "build-time/shared-not-installed" },
+  },
+  "create-plan": {
+    name: "create-plan",
+    variants: ["codex", "claude-code", "cursor", "opencode", "pi"],
+    requiresSuperpowers: ["brainstorming", "writing-plans"],
+    requiresReviewerRuntime: true,
+  },
+  "do-task": {
+    name: "do-task",
+    variants: ["codex", "claude-code", "cursor", "opencode", "pi"],
+    requiresSuperpowers: ["brainstorming", "test-driven-development", "verification-before-completion", "finishing-a-development-branch"],
+    conditionalSuperpowers: ["using-git-worktrees"],
+    requiresReviewerRuntime: true,
+    requiresNotifier: true,
+  },
+  "implement-plan": {
+    name: "implement-plan",
+    variants: ["codex", "claude-code", "cursor", "opencode", "pi"],
+    requiresSuperpowers: ["executing-plans", "using-git-worktrees", "verification-before-completion", "finishing-a-development-branch"],
+    requiresReviewerRuntime: true,
+  },
+  "web-automation": {
+    name: "web-automation",
+    variants: ["codex", "claude-code", "cursor", "opencode", "pi"],
+    requiresSuperpowers: [],
+    requiresReviewerRuntime: false,
+    bootstrap: "web-automation",
+    bootstrapPrerequisites: ["node>=20", "pnpm|corepack", "cloakbrowser"],
+  },
+};
+
+export function expandHome(value, cwd = process.cwd(), homeDir = homedir()) {
+  if (!value) return value;
+  let expanded = value.replace(/^~(?=$|\/)/, homeDir);
+  if (!path.isAbsolute(expanded)) expanded = path.resolve(cwd, expanded);
+  return expanded;
+}
+
+export function resolveClientScope(clientId, scope = "global", cwd = process.cwd()) {
+  const client = CLIENTS[clientId];
+  if (!client) throw new Error(`Unsupported client: ${clientId}`);
+  const scopeConfig = client.scopes[scope];
+  if (!scopeConfig) throw new Error(`Unsupported scope for ${clientId}: ${scope}`);
+  const skillsRoot = expandHome(scopeConfig.skillsRoot, cwd);
+  return { ...scopeConfig, skillsRoot };
+}
+
+export function reviewerRuntimeRoot(clientId, skillsRoot, cwd = process.cwd()) {
+  const template = CLIENTS[clientId].reviewerRuntime.root;
+  return expandHome(template.replace("<skillsRoot>", skillsRoot), cwd);
+}
+
+export function parseReviewerShorthand(value) {
+  if (typeof value !== "string") return null;
+  const slash = value.indexOf("/");
+  if (slash <= 0) return null;
+  if (value.slice(0, slash) !== "pi") return null;
+  const model = value.slice(slash + 1);
+  if (!model) return null;
+  return { reviewerCli: "pi", reviewerModel: model };
+}
+
+export function getSkillSource(skillName, clientId, repoRoot = process.cwd()) {
+  const skill = SKILLS[skillName];
+  const client = CLIENTS[clientId];
+  if (!skill || !client || !skill.variants.includes(client.variant)) return null;
+  if (clientId === "pi") return path.join(repoRoot, "pi-package", "skills", skillName);
+  return path.join(repoRoot, "skills", skillName, client.variant);
+}
+
+export async function pathExists(candidate) {
+  try {
+    await access(candidate, constants.F_OK);
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+export async function detectInstalledClients({ cwd = process.cwd() } = {}) {
+  const result = {};
+  for (const [clientId, client] of Object.entries(CLIENTS)) {
+    let confidence = "not-found";
+    let detail = "not detected";
+    for (const command of client.detectCommands) {
+      const proc = spawnSync(command[0], command.slice(1), { encoding: "utf8" });
+      if (proc.status === 0) {
+        confidence = "cli";
+        detail = (proc.stdout || proc.stderr || "detected").trim().split("\n")[0];
+        break;
+      }
+    }
+    if (confidence === "not-found") {
+      for (const scopeName of Object.keys(client.scopes)) {
+        const { skillsRoot } = resolveClientScope(clientId, scopeName, cwd);
+        if (await pathExists(skillsRoot)) {
+          confidence = "directory";
+          detail = skillsRoot;
+          break;
+        }
+      }
+    }
+    result[clientId] = { clientId, label: client.label, confidence, detail };
+  }
+  return result;
+}
+
+export async function detectInstalledSkills({ skillsRoot, clientId, repoRoot = process.cwd() }) {
+  const state = {};
+  for (const skillName of Object.keys(SKILLS)) {
+    if (getSkillSource(skillName, clientId) === null) {
+      state[skillName] = { state: "unsupported", path: null };
+      continue;
+    }
+    const target = path.join(skillsRoot, skillName);
+    const skillMd = path.join(target, "SKILL.md");
+    let installState = "not-installed";
+    if (existsSync(skillMd)) {
+      installState = "installed";
+      const source = getSkillSource(skillName, clientId, repoRoot);
+      const sourceSkill = source && path.join(source, "SKILL.md");
+      if (sourceSkill && existsSync(sourceSkill)) {
+        const [targetStat, sourceStat] = await Promise.all([stat(skillMd), stat(sourceSkill)]);
+        if (sourceStat.mtimeMs > targetStat.mtimeMs + 1000) installState = "stale";
+      } else {
+        installState = "unknown";
+      }
+    }
+    state[skillName] = {
+      state: installState,
+      path: target,
+    };
+  }
+  return state;
+}
+
+export async function findInstalledSuperpowers(clientId, cwd = process.cwd(), { homeDir = homedir() } = {}) {
+  const client = CLIENTS[clientId];
+  const found = new Set();
+  for (const root of client.superpowers.roots) {
+    const expanded = expandHome(root, cwd, homeDir);
+    if (await pathExists(expanded)) found.add(expanded);
+  }
+  if (clientId === "claude-code") {
+    for (const root of await findClaudeCodeSuperpowersPluginRoots(homeDir)) found.add(root);
+  }
+  if (clientId === "cursor") {
+    for (const root of await findCursorSuperpowersPluginRoots(homeDir)) found.add(root);
+  }
+  return [...found];
+}
+
+async function findClaudeCodeSuperpowersPluginRoots(homeDir) {
+  const pluginId = "superpowers@claude-plugins-official";
+  const settings = await readJsonIfExists(path.join(homeDir, ".claude", "settings.json"));
+  if (settings?.enabledPlugins?.[pluginId] !== true) return [];
+
+  const installed = await readJsonIfExists(path.join(homeDir, ".claude", "plugins", "installed_plugins.json"));
+  const entries = installed?.plugins?.[pluginId];
+  if (!Array.isArray(entries)) return [];
+
+  const roots = [];
+  for (const entry of entries) {
+    if (!entry?.installPath) continue;
+    const skillsRoot = path.join(entry.installPath, "skills");
+    if (await pathExists(skillsRoot)) roots.push(skillsRoot);
+  }
+  return roots;
+}
+
+async function findCursorSuperpowersPluginRoots(homeDir) {
+  const pluginRoot = path.join(homeDir, ".cursor", "plugins", "cache", "cursor-public", "superpowers");
+  let entries = [];
+  try {
+    entries = await readdir(pluginRoot, { withFileTypes: true });
+  } catch (error) {
+    if (error?.code === "ENOENT") return [];
+    throw error;
+  }
+
+  const roots = [];
+  for (const entry of entries) {
+    if (!entry.isDirectory()) continue;
+    const skillsRoot = path.join(pluginRoot, entry.name, "skills");
+    if (await pathExists(skillsRoot)) roots.push(skillsRoot);
+  }
+  return roots.sort();
+}
+
+function piPackageSettingsPath(scope, repoRoot) {
+  if (scope === "packageLocal") return path.join(repoRoot, ".pi", "settings.json");
+  return path.join(homedir(), ".pi", "agent", "settings.json");
+}
+
+async function readJsonIfExists(filePath) {
+  try {
+    return JSON.parse(await readFile(filePath, "utf8"));
+  } catch (error) {
+    if (error?.code === "ENOENT") return null;
+    throw error;
+  }
+}
+
+export async function isPiPackageInstalled({ scope, repoRoot }) {
+  const settingsPath = piPackageSettingsPath(scope, repoRoot);
+  const settings = await readJsonIfExists(settingsPath);
+  if (!Array.isArray(settings?.packages)) return false;
+  const settingsDir = path.dirname(settingsPath);
+  const expected = path.resolve(repoRoot);
+  return settings.packages.some((packagePath) => path.resolve(settingsDir, packagePath) === expected);
+}
+
+export function isBootstrapInstalled(action, target) {
+  const scriptsDir = path.join(target, "scripts");
+  if (action === "pnpm-install") return existsSync(path.join(scriptsDir, "node_modules"));
+  if (action === "web-automation") {
+    return existsSync(path.join(scriptsDir, "node_modules"))
+      && existsSync(path.join(scriptsDir, "node_modules", ".bin", "cloakbrowser"));
+  }
+  return false;
+}
+
+export async function detectHelperInstallState({ source, target, files }) {
+  let differs = false;
+  for (const file of files) {
+    const sourceFile = path.join(source, file);
+    const targetFile = path.join(target, file);
+    if (!existsSync(targetFile)) return "not-installed";
+    if (!existsSync(sourceFile)) return "unknown";
+    const [sourceContents, targetContents] = await Promise.all([readFile(sourceFile), readFile(targetFile)]);
+    if (!sourceContents.equals(targetContents)) differs = true;
+  }
+  return differs ? "stale" : "installed";
+}
+
+export function piPackageCommand({ action, repoRoot, piInstallArg = "" }) {
+  if (["install", "update", "reinstall"].includes(action)) {
+    const args = ["install"];
+    if (piInstallArg) args.push(piInstallArg);
+    args.push(repoRoot);
+    return ["pi", args];
+  }
+  if (action === "remove") {
+    const args = ["remove"];
+    if (piInstallArg) args.push(piInstallArg);
+    args.push(repoRoot);
+    return ["pi", args];
+  }
+  throw new Error(`pi package mode does not support action: ${action}`);
+}
+
+export function requiredSuperpowersFor(actions) {
+  const required = new Set();
+  for (const [skillName, action] of Object.entries(actions || {})) {
+    if (!["install", "update", "reinstall"].includes(action)) continue;
+    for (const skill of SKILLS[skillName]?.requiresSuperpowers || []) required.add(skill);
+  }
+  return [...required];
+}
+
+async function buildReviewerRuntimeOperation({ clientId, scope, skillsRoot, repoRoot, action = null }) {
+  if (action === "skip") return null;
+  const helperTarget = reviewerRuntimeRoot(clientId, skillsRoot, repoRoot);
+  const helperSource = path.join(repoRoot, CLIENTS[clientId].reviewerRuntime.source);
+  const helperFiles = CLIENTS[clientId].reviewerRuntime.files;
+  const helperState = await detectHelperInstallState({ source: helperSource, target: helperTarget, files: helperFiles });
+  const effectiveAction = action || (helperState === "stale" || helperState === "unknown" ? "update" : "install");
+  const operation = {
+    kind: "helper",
+    helper: "reviewer-runtime",
+    clientId,
+    scope,
+    action: effectiveAction,
+    source: helperSource,
+    target: helperTarget,
+    files: helperFiles,
+    skillsRoot,
+  };
+  if (!action && helperState === "installed") {
+    operation.status = "skipped";
+    operation.details = "runtime helper already installed";
+  }
+  return operation;
+}
+
+export async function buildOperationPlan({ selections, repoRoot = process.cwd(), assumeYes = false, superpowersByClient = null }) {
+  const operations = [];
+  const prompts = [];
+  const reportRows = [];
+  const plannedHelpers = new Set();
+
+  for (const selection of selections || []) {
+    const { clientId, scope = "global" } = selection;
+    const client = CLIENTS[clientId];
+    if (!client) throw new Error(`Unsupported client: ${clientId}`);
+    const scopeInfo = selection.skillsRoot ? { skillsRoot: selection.skillsRoot } : resolveClientScope(clientId, scope, repoRoot);
+    const skillsRoot = scopeInfo.skillsRoot;
+    if (clientId === "pi" && scopeInfo.packageMode) {
+      const action = selection.action || "install";
+      if (action === "skip") continue;
+      if (["update", "reinstall"].includes(action)) {
+        operations.push({ kind: "sync-pi-package", clientId, scope, action: "sync", repoRoot });
+      }
+      const packageInstalled = await isPiPackageInstalled({ scope, repoRoot });
+      if (action === "remove") {
+        operations.push({
+          kind: "pi-package",
+          clientId,
+          scope,
+          action,
+          repoRoot,
+          piInstallArg: scopeInfo.piInstallArg || "",
+          status: packageInstalled ? undefined : "skipped",
+          details: packageInstalled ? "" : "not installed in Pi settings",
+        });
+        continue;
+      }
+      operations.push({
+        kind: "pi-package",
+        clientId,
+        scope,
+        action,
+        repoRoot,
+        piInstallArg: scopeInfo.piInstallArg || "",
+        status: packageInstalled && action === "install" ? "skipped" : undefined,
+        details: packageInstalled && action === "install" ? "already installed in Pi settings" : "",
+      });
+      for (const skillName of Object.keys(SKILLS)) {
+        if (getSkillSource(skillName, clientId, repoRoot)) {
+          operations.push({ kind: "package-skill", clientId, scope, skill: skillName, action: "included", status: "included", details: "included in Pi package" });
+        }
+      }
+      for (const bootstrap of [
+        { skill: "atlassian", action: "pnpm-install" },
+        { skill: "web-automation", action: "web-automation" },
+      ]) {
+        const target = path.join(repoRoot, "pi-package", "skills", bootstrap.skill);
+        const installed = isBootstrapInstalled(bootstrap.action, target);
+        const skipBootstrap = action === "install" && installed;
+        operations.push({
+          kind: "bootstrap",
+          clientId,
+          scope,
+          skill: bootstrap.skill,
+          action: bootstrap.action,
+          displayAction: "bootstrap-deps",
+          target,
+          status: skipBootstrap ? "skipped" : undefined,
+          details: skipBootstrap ? "runtime dependencies already installed" : target,
+        });
+      }
+      continue;
+    }
+    const installed = await detectInstalledSkills({ skillsRoot, clientId, repoRoot });
+    const actions = selection.actions || {};
+    const helperActions = selection.helperActions || {};
+    const missingSuperpowers = requiredSuperpowersFor(actions);
+
+    if (missingSuperpowers.length > 0) {
+      const installedSuperpowers = superpowersByClient && Object.hasOwn(superpowersByClient, clientId)
+        ? superpowersByClient[clientId]
+        : await findInstalledSuperpowers(clientId, repoRoot);
+      if (installedSuperpowers.length === 0) {
+        prompts.push({ kind: "missing-superpowers", clientId, scope, required: missingSuperpowers });
+      }
+    }
+
+    for (const [helper, action] of Object.entries(helperActions)) {
+      if (helper !== "reviewer-runtime") continue;
+      const helperTarget = reviewerRuntimeRoot(clientId, skillsRoot, repoRoot);
+      const helperKey = `${clientId}\0${scope}\0${helperTarget}`;
+      if (plannedHelpers.has(helperKey)) continue;
+      plannedHelpers.add(helperKey);
+      const helperOperation = await buildReviewerRuntimeOperation({ clientId, scope, skillsRoot, repoRoot, action });
+      if (helperOperation) operations.push(helperOperation);
+    }
+
+    for (const [skillName, action] of Object.entries(actions)) {
+      const source = getSkillSource(skillName, clientId, repoRoot);
+      if (!source) {
+        operations.push({ kind: "skill", clientId, scope, skill: skillName, action: "unsupported", status: "skipped", details: "variant unavailable" });
+        continue;
+      }
+      if (action === "skip") continue;
+      const target = path.join(skillsRoot, skillName);
+      if (clientId === "pi" && !operations.some((op) => op.kind === "sync-pi-package")) {
+        operations.push({ kind: "sync-pi-package", clientId, scope, action: "sync", repoRoot });
+      }
+      operations.push({ kind: "skill", clientId, scope, skill: skillName, action, source, target, skillsRoot });
+      if (["install", "update", "reinstall"].includes(action) && SKILLS[skillName].requiresReviewerRuntime) {
+        const helperTarget = reviewerRuntimeRoot(clientId, skillsRoot, repoRoot);
+        const helperKey = `${clientId}\0${scope}\0${helperTarget}`;
+        if (!plannedHelpers.has(helperKey)) {
+          plannedHelpers.add(helperKey);
+          const helperOperation = await buildReviewerRuntimeOperation({ clientId, scope, skillsRoot, repoRoot });
+          if (helperOperation) operations.push(helperOperation);
+        }
+      }
+      if (["install", "update", "reinstall"].includes(action) && SKILLS[skillName].bootstrap) {
+        operations.push({ kind: "bootstrap", clientId, scope, skill: skillName, item: `${skillName} deps`, action: SKILLS[skillName].bootstrap, displayAction: "bootstrap-deps", target });
+      }
+    }
+
+    const workflowSkills = Object.keys(SKILLS).filter((skillName) => (SKILLS[skillName].requiresSuperpowers || []).length > 0);
+    const removing = workflowSkills.filter((skillName) => actions[skillName] === "remove");
+    if (removing.length > 0) {
+      const remaining = workflowSkills.some((skillName) => actions[skillName] !== "remove" && installed[skillName]?.state === "installed");
+      if (!remaining) prompts.push({ kind: "remove-superpowers", clientId, scope, removedWorkflowSkills: removing });
+    }
+  }
+
+  for (const op of operations) {
+    reportRows.push({
+      client: op.clientId,
+      scope: op.scope,
+      item: op.item || op.skill || op.helper || op.kind,
+      action: op.displayAction || op.action,
+      status: op.status || "planned",
+      details: op.details || op.target || "",
+    });
+  }
+
+  return { operations, prompts, reportRows, assumeYes };
+}
+
+export async function validateRemoveTarget(target, skillsRoot, { repoRoot = process.cwd() } = {}) {
+  const resolvedRoot = path.resolve(skillsRoot);
+  const resolvedTarget = path.resolve(target);
+  const home = path.resolve(homedir());
+  const resolvedRepo = path.resolve(repoRoot);
+  if (resolvedTarget === resolvedRoot) throw new Error("refusing to remove skills root itself");
+  if (resolvedTarget === home) throw new Error("refusing to remove home directory");
+  if (resolvedTarget === resolvedRepo) throw new Error("refusing to remove repository root");
+  if (resolvedTarget === path.parse(resolvedTarget).root) throw new Error("refusing to remove filesystem root");
+  const relative = path.relative(resolvedRoot, resolvedTarget);
+  if (relative.startsWith("..") || path.isAbsolute(relative) || relative === "") {
+    throw new Error(`refusing to remove target outside skills root: ${target}`);
+  }
+  try {
+    const info = await lstat(resolvedTarget);
+    if (!info.isSymbolicLink()) {
+      const [realRoot, realTarget] = await Promise.all([realpath(resolvedRoot), realpath(resolvedTarget)]);
+      const realRelative = path.relative(realRoot, realTarget);
+      if (realRelative.startsWith("..") || path.isAbsolute(realRelative) || realRelative === "") {
+        throw new Error(`refusing to remove real target outside skills root: ${target}`);
+      }
+    }
+  } catch (error) {
+    if (error && error.code !== "ENOENT") throw error;
+  }
+  return true;
+}
+
+export async function copyDirectoryReplacing(source, target) {
+  await mkdir(path.dirname(target), { recursive: true });
+  await rm(target, { recursive: true, force: true });
+  await cp(source, target, { recursive: true, force: true, dereference: false });
+}
+
+export async function installHelperAllowlist({ source, target, files }) {
+  await mkdir(target, { recursive: true });
+  for (const file of files) {
+    await cp(path.join(source, file), path.join(target, file), { force: true });
+    if (file.endsWith(".sh")) await chmod(path.join(target, file), 0o755);
+  }
+}
+
+function runCommand(command, args, options = {}) {
+  const result = spawnSync(command, args, { encoding: "utf8", stdio: "pipe", ...options });
+  if (result.status !== 0) {
+    const detail = (result.stderr || result.stdout || `${command} exited ${result.status}`).trim();
+    throw new Error(detail);
+  }
+  return (result.stdout || result.stderr || "ok").trim();
+}
+
+function resolvePnpmCommand() {
+  if (spawnSync("pnpm", ["--version"], { encoding: "utf8" }).status === 0) return ["pnpm"];
+  if (spawnSync("corepack", ["--version"], { encoding: "utf8" }).status === 0) return ["corepack", "pnpm"];
+  throw new Error("missing prerequisite: pnpm or corepack");
+}
+
+function requireNode20() {
+  const major = Number.parseInt(process.versions.node.split(".")[0], 10);
+  if (major < 20) throw new Error(`missing prerequisite: Node.js 20+ (found ${process.version})`);
+}
+
+export async function executeOperation(op) {
+  if (op.action === "unsupported" || op.status === "skipped") return { ...op, status: "skipped" };
+  if (op.kind === "package-skill") return { ...op, status: "included" };
+  if (op.kind === "sync-pi-package") {
+    runCommand(path.join(op.repoRoot, "scripts", "sync-pi-package-skills.sh"), [], { cwd: op.repoRoot });
+    return { ...op, status: "ok" };
+  }
+  if (op.kind === "pi-package") {
+    const [command, args] = piPackageCommand(op);
+    runCommand(command, args, { cwd: op.repoRoot });
+    return { ...op, status: "ok" };
+  }
+  if (op.kind === "skill") {
+    if (op.action === "remove") {
+      await validateRemoveTarget(op.target, op.skillsRoot, { repoRoot: REPO_ROOT });
+      const info = existsSync(op.target) ? await lstat(op.target) : null;
+      if (info?.isSymbolicLink()) await unlink(op.target);
+      else await rm(op.target, { recursive: true, force: true });
+      return { ...op, status: "ok" };
+    }
+    await copyDirectoryReplacing(op.source, op.target);
+    return { ...op, status: "ok" };
+  }
+  if (op.kind === "helper") {
+    if (op.action === "remove") {
+      await validateRemoveTarget(op.target, op.skillsRoot, { repoRoot: REPO_ROOT });
+      const info = existsSync(op.target) ? await lstat(op.target) : null;
+      if (info?.isSymbolicLink()) await unlink(op.target);
+      else await rm(op.target, { recursive: true, force: true });
+      return { ...op, status: "ok" };
+    }
+    await installHelperAllowlist(op);
+    return { ...op, status: "ok" };
+  }
+  if (op.kind === "superpowers") {
+    await mkdir(path.dirname(op.target), { recursive: true });
+    if (op.action === "remove") {
+      await validateRemoveTarget(op.target, op.skillsRoot, { repoRoot: REPO_ROOT });
+      await rm(op.target, { recursive: true, force: true });
+    } else if (op.mode === "copy") {
+      await copyDirectoryReplacing(op.source, op.target);
+    } else {
+      await rm(op.target, { recursive: true, force: true });
+      await symlink(op.source, op.target, "dir");
+    }
+    return { ...op, status: "ok" };
+  }
+  if (op.kind === "bootstrap") {
+    requireNode20();
+    const pnpm = resolvePnpmCommand();
+    const scriptsDir = path.join(op.target, "scripts");
+    if (op.action === "pnpm-install") {
+      runCommand(pnpm[0], [...pnpm.slice(1), "install", "--frozen-lockfile", "--dir", scriptsDir]);
+    } else if (op.action === "web-automation") {
+      runCommand(pnpm[0], [...pnpm.slice(1), "install", "--frozen-lockfile", "--dir", scriptsDir]);
+      runCommand(pnpm[0], [...pnpm.slice(1), "--dir", scriptsDir, "exec", "cloakbrowser", "install"]);
+      runCommand(pnpm[0], [...pnpm.slice(1), "rebuild", "--dir", scriptsDir, "better-sqlite3", "esbuild"]);
+    }
+    return { ...op, status: "ok" };
+  }
+  return { ...op, status: "warning", details: "operation is planned/manual" };
+}

scripts/manage-skills.mjs

@@ -0,0 +1,320 @@
+#!/usr/bin/env node
+import { readFile } from "node:fs/promises";
+import { stdin as input, stdout as output } from "node:process";
+import readline from "node:readline/promises";
+import path from "node:path";
+
+import {
+  CLIENTS,
+  SKILLS,
+  buildOperationPlan,
+  detectHelperInstallState,
+  detectInstalledClients,
+  detectInstalledSkills,
+  executeOperation,
+  isPiPackageInstalled,
+  parseReviewerShorthand,
+  resolveClientScope,
+  reviewerRuntimeRoot,
+} from "./lib/skill-manager-core.mjs";
+
+function usage() {
+  return `Usage:
+  node scripts/manage-skills.mjs [--dry-run]
+  node scripts/manage-skills.mjs --plan-only --answers <answers.json>
+  node scripts/manage-skills.mjs --client <client> --scope <scope> --skill <skill> --action <install|update|reinstall|remove|skip> [--yes]
+
+Options:
+  --dry-run              Prompt or detect normally, but do not modify files.
+  --plan-only            Non-interactive mode. Requires --answers and emits JSON.
+  --answers <file>       JSON answers file with { "selections": [...] }.
+  --client <client>      codex, claude-code, cursor, opencode, or pi.
+  --scope <scope>        Client scope. Examples: global, local, packageGlobal, packageLocal.
+  --skill <skill>        Skill to manage. May be repeated.
+  --action <action>      Action for --skill entries. Defaults to install.
+  --yes                  Execute planned operations without final confirmation.
+  --pi-package           With --client pi, default to packageGlobal scope (full bundle).
+  --reviewer <value>     Parse/display reviewer shorthand, e.g. pi/claude-opus-4-7.
+  -h, --help             Show this help.
+
+Answers JSON example:
+{
+  "selections": [
+    {
+      "clientId": "codex",
+      "scope": "global",
+      "actions": { "create-plan": "install", "web-automation": "skip" },
+      "helperActions": { "reviewer-runtime": "skip" }
+    }
+  ]
+}
+
+Final report columns: client, scope, skill/helper, action, status, details.
+`;
+}
+
+function parseArgs(argv) {
+  const args = { skills: [] };
+  for (let i = 0; i < argv.length; i += 1) {
+    const arg = argv[i];
+    switch (arg) {
+      case "--dry-run": args.dryRun = true; break;
+      case "--plan-only": args.planOnly = true; break;
+      case "--answers": args.answers = argv[++i]; break;
+      case "--client": args.client = argv[++i]; break;
+      case "--scope": args.scope = argv[++i]; break;
+      case "--skill": args.skills.push(argv[++i]); break;
+      case "--action": args.action = argv[++i]; break;
+      case "--yes": args.yes = true; break;
+      case "--pi-package": args.piPackage = true; break;
+      case "--reviewer": args.reviewer = argv[++i]; break;
+      case "-h":
+      case "--help": args.help = true; break;
+      default:
+        throw new Error(`Unknown argument: ${arg}`);
+    }
+  }
+  return args;
+}
+
+function renderTable(rows) {
+  if (!rows.length) return "No operations planned.";
+  const headers = ["client", "scope", "skill/helper", "action", "status", "details"];
+  const data = rows.map((row) => [row.client, row.scope, row.item, row.action, row.status, row.details]);
+  const widths = headers.map((header, index) => Math.max(header.length, ...data.map((row) => String(row[index] ?? "").length)));
+  const format = (row) => row.map((cell, index) => String(cell ?? "").padEnd(widths[index])).join("  ");
+  return [format(headers), format(widths.map((width) => "-".repeat(width))), ...data.map(format)].join("\n");
+}
+
+async function answersToPlan(answers, { assumeYes = false } = {}) {
+  const plan = await buildOperationPlan({ selections: answers.selections || [], assumeYes, repoRoot: process.cwd(), superpowersByClient: answers.superpowersByClient });
+  return plan;
+}
+
+async function buildCliSelection(args) {
+  if (!args.client) return null;
+  const clientId = args.client;
+  if (!CLIENTS[clientId]) throw new Error(`Unsupported client: ${clientId}`);
+  const scope = args.scope || (clientId === "pi" && args.piPackage ? "packageGlobal" : "global");
+  const scopeInfo = resolveClientScope(clientId, scope, process.cwd());
+  if (clientId === "pi" && scopeInfo.packageMode) {
+    return { selections: [{ clientId, scope, action: args.action || "install", actions: {} }] };
+  }
+  const skills = args.skills.length ? args.skills : Object.keys(SKILLS);
+  const action = args.action || "install";
+  const actions = Object.fromEntries(skills.map((skill) => [skill, action]));
+  return { selections: [{ clientId, scope, actions }] };
+}
+
+async function interactiveAnswers({ dryRun = false } = {}) {
+  const detected = await detectInstalledClients();
+  console.log("Detected supported clients:");
+  for (const info of Object.values(detected)) {
+    console.log(`- ${info.clientId}: ${info.confidence} (${info.detail})`);
+  }
+  if (dryRun && !input.isTTY) {
+    return { selections: [] };
+  }
+  const rl = readline.createInterface({ input, output });
+  try {
+    const clientLine = await rl.question("Clients to manage (comma-separated, blank for detected CLI clients): ");
+    const chosenClients = clientLine.trim()
+      ? clientLine.split(",").map((value) => value.trim()).filter(Boolean)
+      : Object.values(detected).filter((info) => info.confidence !== "not-found").map((info) => info.clientId);
+    const selections = [];
+    for (const clientId of chosenClients) {
+      if (!CLIENTS[clientId]) {
+        console.log(`Skipping unsupported client: ${clientId}`);
+        continue;
+      }
+      const scopeNames = Object.keys(CLIENTS[clientId].scopes);
+      let scope = scopeNames[0];
+      if (scopeNames.length > 1) {
+        const answer = await rl.question(`${clientId} scope (${scopeNames.join("/")}) [${scope}]: `);
+        if (answer.trim()) scope = answer.trim();
+      }
+      const scopeInfo = resolveClientScope(clientId, scope, process.cwd());
+      if (scopeInfo.packageMode) {
+        console.log(`${clientId}/${scope}: package mode manages the full Pi package bundle; per-skill prompts are skipped.`);
+        const installed = await isPiPackageInstalled({ scope, repoRoot: process.cwd() });
+        const choices = "install/update/reinstall/remove/skip";
+        const defaultAction = installed ? "skip" : "install";
+        const answer = await rl.question(`${clientId}/${scope} package is ${installed ? "installed" : "not-installed"}; action (${choices}) [${defaultAction}]: `);
+        const chosen = answer.trim() || defaultAction;
+        if (!choices.split("/").includes(chosen)) {
+          console.log(`Invalid action '${chosen}', using skip.`);
+          selections.push({ clientId, scope, action: "skip", actions: {} });
+        } else {
+          selections.push({ clientId, scope, action: chosen, actions: {} });
+        }
+        continue;
+      }
+      const installed = await detectInstalledSkills({ clientId, skillsRoot: scopeInfo.skillsRoot });
+      const actions = {};
+      for (const skill of Object.keys(SKILLS)) {
+        const state = installed[skill]?.state || "unsupported";
+        if (state === "unsupported") {
+          console.log(`${clientId}/${scope}/${skill}: unsupported`);
+          actions[skill] = "skip";
+          continue;
+        }
+        const choices = state === "installed" || state === "stale" || state === "unknown" ? "update/reinstall/remove/skip" : "install/skip";
+        const defaultAction = "skip";
+        const answer = await rl.question(`${clientId}/${scope}/${skill} is ${state}; action (${choices}) [${defaultAction}]: `);
+        const chosen = answer.trim() || defaultAction;
+        const allowed = choices.split("/");
+        if (!allowed.includes(chosen)) {
+          console.log(`Invalid action '${chosen}', using skip.`);
+          actions[skill] = "skip";
+        } else {
+          actions[skill] = chosen;
+        }
+      }
+      const helperActions = {};
+      const workflowNeedsReviewerRuntime = Object.entries(actions).some(([skill, action]) => (
+        ["install", "update", "reinstall"].includes(action) && SKILLS[skill]?.requiresReviewerRuntime
+      ));
+      const helperTarget = reviewerRuntimeRoot(clientId, scopeInfo.skillsRoot, process.cwd());
+      const helperState = await detectHelperInstallState({
+        source: path.join(process.cwd(), CLIENTS[clientId].reviewerRuntime.source),
+        target: helperTarget,
+        files: CLIENTS[clientId].reviewerRuntime.files,
+      });
+      if (workflowNeedsReviewerRuntime || helperState !== "not-installed") {
+        const choices = helperState === "not-installed" ? "install/skip" : "update/reinstall/remove/skip";
+        let defaultAction = "skip";
+        if (workflowNeedsReviewerRuntime && helperState === "not-installed") defaultAction = "install";
+        if (workflowNeedsReviewerRuntime && ["stale", "unknown"].includes(helperState)) defaultAction = "update";
+        const answer = await rl.question(`${clientId}/${scope}/reviewer-runtime is ${helperState}; action (${choices}) [${defaultAction}]: `);
+        const chosen = answer.trim() || defaultAction;
+        const allowed = choices.split("/");
+        if (!allowed.includes(chosen)) {
+          console.log(`Invalid action '${chosen}', using skip.`);
+          helperActions["reviewer-runtime"] = "skip";
+        } else {
+          helperActions["reviewer-runtime"] = chosen;
+        }
+      }
+      selections.push({ clientId, scope, actions, helperActions });
+    }
+    return { selections };
+  } finally {
+    rl.close();
+  }
+}
+
+async function main() {
+  const args = parseArgs(process.argv.slice(2));
+  if (args.help) {
+    console.log(usage());
+    return;
+  }
+  if (args.reviewer) {
+    console.log(JSON.stringify(parseReviewerShorthand(args.reviewer), null, 2));
+    return;
+  }
+
+  let answers;
+  if (args.answers) {
+    answers = JSON.parse(await readFile(path.resolve(args.answers), "utf8"));
+  } else {
+    answers = await buildCliSelection(args);
+  }
+  if (!answers) answers = await interactiveAnswers({ dryRun: args.dryRun });
+
+  const plan = await answersToPlan(answers, { assumeYes: args.yes });
+
+  if (args.planOnly) {
+    console.log(JSON.stringify(plan, null, 2));
+    return;
+  }
+
+  console.log("Operation preview:");
+  console.log(renderTable(plan.reportRows));
+  if (plan.prompts.length) {
+    console.log("\nDependency prompts:");
+    for (const prompt of plan.prompts) {
+      const promptDetail = prompt.required ? prompt.required.join(",") : (prompt.removedWorkflowSkills ? prompt.removedWorkflowSkills.join(",") : "");
+      console.log(`- ${prompt.kind}: ${prompt.clientId}/${prompt.scope} ${promptDetail}`);
+    }
+  }
+
+  if (plan.operations.length === 0) return;
+
+  if (args.dryRun) {
+    console.log("\nDry-run mode: no filesystem changes performed.");
+    return;
+  }
+
+  if (plan.prompts.length && input.isTTY && !args.yes) {
+    const rl = readline.createInterface({ input, output });
+    try {
+      for (const prompt of plan.prompts) {
+        if (prompt.kind === "missing-superpowers") {
+          const install = await rl.question(`Install/link Superpowers for ${prompt.clientId}/${prompt.scope}? (yes/no) [no]: `);
+          if (install.trim().toLowerCase() === "yes") {
+            const source = await rl.question("Absolute path to Obra Superpowers skills directory: ");
+            const modeAnswer = await rl.question("Install mode (symlink/copy) [symlink]: ");
+            const client = CLIENTS[prompt.clientId];
+            const scope = resolveClientScope(prompt.clientId, prompt.scope, process.cwd());
+            const target = client.superpowers.layout === "cursor-nested"
+              ? `${scope.skillsRoot}/superpowers/skills`
+              : `${scope.skillsRoot}/superpowers`;
+            plan.operations.push({ kind: "superpowers", clientId: prompt.clientId, scope: prompt.scope, action: "install", source: source.trim(), target, skillsRoot: scope.skillsRoot, mode: modeAnswer.trim() === "copy" ? "copy" : "symlink" });
+          }
+        }
+        if (prompt.kind === "remove-superpowers") {
+          const removeAnswer = await rl.question(`Remove Superpowers for ${prompt.clientId}/${prompt.scope} too? (yes/no) [no]: `);
+          if (removeAnswer.trim().toLowerCase() === "yes") {
+            const scope = resolveClientScope(prompt.clientId, prompt.scope, process.cwd());
+            const client = CLIENTS[prompt.clientId];
+            const target = `${scope.skillsRoot}/superpowers`;
+            plan.operations.push({ kind: "superpowers", clientId: prompt.clientId, scope: prompt.scope, action: "remove", target, skillsRoot: scope.skillsRoot });
+          }
+        }
+      }
+    } finally {
+      rl.close();
+    }
+  }
+
+  if (!args.yes) {
+    if (!input.isTTY) {
+      console.log("Refusing to execute without --yes in non-interactive mode.");
+      process.exitCode = 2;
+      return;
+    }
+    const rl = readline.createInterface({ input, output });
+    const answer = await rl.question("Proceed with these operations? (yes/no): ");
+    rl.close();
+    if (answer.trim().toLowerCase() !== "yes") {
+      console.log("Skipped by user.");
+      return;
+    }
+  }
+
+  const results = [];
+  for (const operation of plan.operations) {
+    try {
+      results.push(await executeOperation(operation));
+    } catch (error) {
+      results.push({ ...operation, status: "failed", details: error.message });
+    }
+  }
+  const rows = results.map((op) => ({
+    client: op.clientId,
+    scope: op.scope,
+    item: op.item || op.skill || op.helper || op.kind,
+    action: op.displayAction || op.action,
+    status: op.status,
+    details: op.details || op.target || "",
+  }));
+  console.log("\nFinal report:");
+  console.log(renderTable(rows));
+  if (results.some((result) => result.status === "failed")) process.exitCode = 1;
+}
+
+main().catch((error) => {
+  console.error(error.message);
+  process.exitCode = 1;
+});

scripts/manage-skills.sh

@@ -0,0 +1,3 @@
+#!/usr/bin/env bash
+set -euo pipefail
+exec node "$(dirname "$0")/manage-skills.mjs" "$@"

scripts/tests/skill-manager-core.test.mjs

@@ -0,0 +1,429 @@
+import assert from "node:assert/strict";
+import { execFileSync } from "node:child_process";
+import { mkdtemp, mkdir, writeFile, rm } from "node:fs/promises";
+import { tmpdir } from "node:os";
+import path from "node:path";
+import test from "node:test";
+import { fileURLToPath } from "node:url";
+
+import {
+  CLIENTS,
+  SKILLS,
+  buildOperationPlan,
+  detectInstalledSkills,
+  findInstalledSuperpowers,
+  getSkillSource,
+  piPackageCommand,
+  parseReviewerShorthand,
+  validateRemoveTarget,
+} from "../lib/skill-manager-core.mjs";
+
+const REPO_ROOT = path.resolve(path.dirname(fileURLToPath(import.meta.url)), "../..");
+
+test("manifest records supported variants and helper allowlists", () => {
+  assert.deepEqual(SKILLS["web-automation"].variants, ["codex", "claude-code", "cursor", "opencode", "pi"]);
+  assert.deepEqual(CLIENTS.codex.reviewerRuntime.files, ["run-review.sh", "notify-telegram.sh"]);
+  assert.deepEqual(CLIENTS.pi.reviewerRuntime.files, ["run-review.sh", "notify-telegram.sh"]);
+});
+
+test("parseReviewerShorthand keeps provider-qualified model ids verbatim", () => {
+  assert.deepEqual(parseReviewerShorthand("pi/claude-opus-4-7"), {
+    reviewerCli: "pi",
+    reviewerModel: "claude-opus-4-7",
+  });
+  assert.deepEqual(parseReviewerShorthand("pi/anthropic/claude-opus-4-7"), {
+    reviewerCli: "pi",
+    reviewerModel: "anthropic/claude-opus-4-7",
+  });
+  assert.equal(parseReviewerShorthand("codex/gpt-5"), null);
+});
+
+test("unsupported skill variant is reported as unsupported", () => {
+  assert.ok(getSkillSource("web-automation", "cursor").endsWith("skills/web-automation/cursor"));
+  assert.ok(getSkillSource("web-automation", "pi").endsWith("pi-package/skills/web-automation"));
+});
+
+test("detectInstalledSkills reports installed and missing skills", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-detect-"));
+  try {
+    await mkdir(path.join(dir, "skills", "create-plan"), { recursive: true });
+    await writeFile(path.join(dir, "skills", "create-plan", "SKILL.md"), "---\nname: create-plan\n---\n");
+    const state = await detectInstalledSkills({ skillsRoot: path.join(dir, "skills"), clientId: "codex" });
+    assert.equal(state["create-plan"].state, "installed");
+    assert.equal(state["web-automation"].state, "not-installed");
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("plan install workflow skill includes reviewer-runtime and missing Superpowers prompt", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-plan-"));
+  try {
+    await mkdir(path.join(dir, "repo", "skills", "create-plan", "codex"), { recursive: true });
+    await writeFile(path.join(dir, "repo", "skills", "create-plan", "codex", "SKILL.md"), "---\nname: create-plan\n---\n");
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "codex", scope: "global", skillsRoot: path.join(dir, "install"), actions: { "create-plan": "install" } }],
+      assumeYes: true,
+      repoRoot: path.join(dir, "repo"),
+      superpowersByClient: { codex: [] },
+    });
+    assert.equal(plan.operations.some((op) => op.kind === "skill" && op.skill === "create-plan" && op.action === "install"), true);
+    assert.equal(plan.operations.some((op) => op.kind === "helper" && op.helper === "reviewer-runtime"), true);
+    assert.equal(plan.prompts.some((prompt) => prompt.kind === "missing-superpowers" && prompt.clientId === "codex"), true);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("plan skips already current reviewer-runtime helper for workflow skill updates", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-helper-current-"));
+  try {
+    const repo = path.join(dir, "repo");
+    const install = path.join(dir, "install");
+    await mkdir(path.join(repo, "skills", "create-plan", "cursor"), { recursive: true });
+    await mkdir(path.join(repo, "skills", "do-task", "cursor"), { recursive: true });
+    await writeFile(path.join(repo, "skills", "create-plan", "cursor", "SKILL.md"), "---\nname: create-plan\n---\n");
+    await writeFile(path.join(repo, "skills", "do-task", "cursor", "SKILL.md"), "---\nname: do-task\n---\n");
+    await mkdir(path.join(repo, "skills", "reviewer-runtime"), { recursive: true });
+    await mkdir(path.join(install, "reviewer-runtime"), { recursive: true });
+    for (const file of CLIENTS.cursor.reviewerRuntime.files) {
+      await writeFile(path.join(repo, "skills", "reviewer-runtime", file), `${file}\n`);
+      await writeFile(path.join(install, "reviewer-runtime", file), `${file}\n`);
+    }
+
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "cursor", scope: "global", skillsRoot: install, actions: { "create-plan": "update", "do-task": "update" } }],
+      repoRoot: repo,
+      superpowersByClient: { cursor: [path.join(dir, "superpowers")] },
+    });
+
+    const helperRows = plan.reportRows.filter((row) => row.item === "reviewer-runtime");
+    assert.equal(helperRows.length, 1);
+    assert.equal(helperRows[0].action, "install");
+    assert.equal(helperRows[0].status, "skipped");
+    assert.match(helperRows[0].details, /already installed/);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("plan auto-updates stale reviewer-runtime helper for workflow skill updates", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-helper-stale-"));
+  try {
+    const repo = path.join(dir, "repo");
+    const install = path.join(dir, "install");
+    await mkdir(path.join(repo, "skills", "create-plan", "cursor"), { recursive: true });
+    await writeFile(path.join(repo, "skills", "create-plan", "cursor", "SKILL.md"), "---\nname: create-plan\n---\n");
+    await mkdir(path.join(repo, "skills", "reviewer-runtime"), { recursive: true });
+    await mkdir(path.join(install, "reviewer-runtime"), { recursive: true });
+    for (const file of CLIENTS.cursor.reviewerRuntime.files) {
+      await writeFile(path.join(repo, "skills", "reviewer-runtime", file), `${file}:new\n`);
+      await writeFile(path.join(install, "reviewer-runtime", file), `${file}:old\n`);
+    }
+
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "cursor", scope: "global", skillsRoot: install, actions: { "create-plan": "update" } }],
+      repoRoot: repo,
+      superpowersByClient: { cursor: [path.join(dir, "superpowers")] },
+    });
+
+    const helperRows = plan.reportRows.filter((row) => row.item === "reviewer-runtime");
+    assert.equal(helperRows.length, 1);
+    assert.equal(helperRows[0].action, "update");
+    assert.equal(helperRows[0].status, "planned");
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("plan honors explicit reviewer-runtime helper actions", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-helper-explicit-"));
+  try {
+    const repo = path.join(dir, "repo");
+    const install = path.join(dir, "install");
+    await mkdir(path.join(repo, "skills", "reviewer-runtime"), { recursive: true });
+    await mkdir(path.join(install, "reviewer-runtime"), { recursive: true });
+    for (const file of CLIENTS.cursor.reviewerRuntime.files) {
+      await writeFile(path.join(repo, "skills", "reviewer-runtime", file), `${file}\n`);
+      await writeFile(path.join(install, "reviewer-runtime", file), `${file}\n`);
+    }
+
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "cursor", scope: "global", skillsRoot: install, actions: {}, helperActions: { "reviewer-runtime": "reinstall" } }],
+      repoRoot: repo,
+    });
+
+    assert.deepEqual(plan.reportRows.map((row) => [row.item, row.action, row.status]), [
+      ["reviewer-runtime", "reinstall", "planned"],
+    ]);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("plan labels skill bootstrap rows as dependency rows", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-bootstrap-label-"));
+  try {
+    const repo = path.join(dir, "repo");
+    const install = path.join(dir, "install");
+    await mkdir(path.join(repo, "skills", "web-automation", "claude-code"), { recursive: true });
+    await writeFile(path.join(repo, "skills", "web-automation", "claude-code", "SKILL.md"), "---\nname: web-automation\n---\n");
+
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "claude-code", scope: "global", skillsRoot: install, actions: { "web-automation": "update" } }],
+      repoRoot: repo,
+    });
+
+    assert.deepEqual(plan.reportRows.map((row) => [row.item, row.action]), [
+      ["web-automation", "update"],
+      ["web-automation deps", "bootstrap-deps"],
+    ]);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("findInstalledSuperpowers detects Claude Code Superpowers plugin installs", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-claude-superpowers-"));
+  try {
+    const installPath = path.join(dir, ".claude", "plugins", "cache", "claude-plugins-official", "superpowers", "4.2.0");
+    await mkdir(path.join(installPath, "skills", "brainstorming"), { recursive: true });
+    await writeFile(path.join(installPath, "skills", "brainstorming", "SKILL.md"), "---\nname: brainstorming\n---\n");
+    await mkdir(path.join(dir, ".claude", "plugins"), { recursive: true });
+    await writeFile(path.join(dir, ".claude", "settings.json"), JSON.stringify({
+      enabledPlugins: {
+        "superpowers@claude-plugins-official": true,
+      },
+    }));
+    await writeFile(path.join(dir, ".claude", "plugins", "installed_plugins.json"), JSON.stringify({
+      plugins: {
+        "superpowers@claude-plugins-official": [
+          {
+            scope: "user",
+            installPath,
+            version: "4.2.0",
+          },
+        ],
+      },
+    }));
+
+    assert.deepEqual(await findInstalledSuperpowers("claude-code", process.cwd(), { homeDir: dir }), [
+      path.join(installPath, "skills"),
+    ]);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("findInstalledSuperpowers detects OpenCode shared agents Superpowers installs", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-opencode-superpowers-"));
+  try {
+    const sharedRoot = path.join(dir, ".agents", "skills", "superpowers");
+    await mkdir(path.join(sharedRoot, "brainstorming"), { recursive: true });
+    await writeFile(path.join(sharedRoot, "brainstorming", "SKILL.md"), "---\nname: brainstorming\n---\n");
+
+    assert.deepEqual(await findInstalledSuperpowers("opencode", process.cwd(), { homeDir: dir }), [
+      sharedRoot,
+    ]);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("findInstalledSuperpowers detects Cursor Superpowers plugin installs", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-cursor-superpowers-"));
+  try {
+    const pluginSkills = path.join(dir, ".cursor", "plugins", "cache", "cursor-public", "superpowers", "abc123", "skills");
+    await mkdir(path.join(pluginSkills, "brainstorming"), { recursive: true });
+    await writeFile(path.join(pluginSkills, "brainstorming", "SKILL.md"), "---\nname: brainstorming\n---\n");
+
+    assert.deepEqual(await findInstalledSuperpowers("cursor", process.cwd(), { homeDir: dir }), [
+      pluginSkills,
+    ]);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("plan removing last workflow skill prompts for optional Superpowers removal", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-remove-"));
+  try {
+    await mkdir(path.join(dir, "skills", "do-task"), { recursive: true });
+    await writeFile(path.join(dir, "skills", "do-task", "SKILL.md"), "---\nname: do-task\n---\n");
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "codex", scope: "global", skillsRoot: path.join(dir, "skills"), actions: { "do-task": "remove" } }],
+      assumeYes: true,
+      repoRoot: process.cwd(),
+    });
+    assert.equal(plan.prompts.some((prompt) => prompt.kind === "remove-superpowers" && prompt.clientId === "codex"), true);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("pi package mode plans full package install instead of per-skill copy", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-pi-package-"));
+  try {
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "pi", scope: "packageLocal", action: "install", actions: { "create-plan": "skip", atlassian: "remove" } }],
+      repoRoot: dir,
+    });
+    assert.equal(plan.operations.some((op) => op.kind === "pi-package" && op.piInstallArg === "-l"), true);
+    assert.equal(plan.operations.some((op) => op.kind === "skill"), false);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("pi package mode surfaces bundled skills and skips already installed package resources", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-pi-package-installed-"));
+  try {
+    const repo = path.join(dir, "repo");
+    await mkdir(path.join(repo, ".pi"), { recursive: true });
+    await writeFile(path.join(repo, ".pi", "settings.json"), JSON.stringify({ packages: [".."] }));
+    for (const skill of Object.keys(SKILLS)) {
+      await mkdir(path.join(repo, "pi-package", "skills", skill), { recursive: true });
+    }
+    await mkdir(path.join(repo, "pi-package", "skills", "atlassian", "scripts", "node_modules"), { recursive: true });
+    await mkdir(path.join(repo, "pi-package", "skills", "web-automation", "scripts", "node_modules", ".bin"), { recursive: true });
+    await writeFile(path.join(repo, "pi-package", "skills", "web-automation", "scripts", "node_modules", ".bin", "cloakbrowser"), "");
+
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "pi", scope: "packageLocal", action: "install", actions: {} }],
+      repoRoot: repo,
+    });
+
+    const packageInstall = plan.operations.find((op) => op.kind === "pi-package");
+    assert.equal(packageInstall.status, "skipped");
+    assert.match(packageInstall.details, /already installed/);
+    assert.deepEqual(
+      plan.reportRows.filter((row) => row.action === "included").map((row) => row.item).sort(),
+      Object.keys(SKILLS).sort()
+    );
+    assert.deepEqual(
+      plan.reportRows.filter((row) => row.action === "bootstrap-deps").map((row) => [row.item, row.status]).sort(),
+      [["atlassian", "skipped"], ["web-automation", "skipped"]]
+    );
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("pi package mode remove skips when package is not installed", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-pi-package-remove-"));
+  try {
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "pi", scope: "packageLocal", action: "remove", actions: {} }],
+      repoRoot: dir,
+    });
+
+    assert.deepEqual(plan.operations.map((op) => op.kind), ["pi-package"]);
+    assert.equal(plan.operations[0].action, "remove");
+    assert.equal(plan.operations[0].status, "skipped");
+    assert.match(plan.operations[0].details, /not installed/);
+    assert.equal(plan.reportRows[0].item, "pi-package");
+    assert.equal(plan.reportRows[0].action, "remove");
+    assert.equal(plan.reportRows[0].status, "skipped");
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("pi package mode remove plans removal when package is installed", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-pi-package-remove-installed-"));
+  try {
+    const repo = path.join(dir, "repo");
+    await mkdir(path.join(repo, ".pi"), { recursive: true });
+    await writeFile(path.join(repo, ".pi", "settings.json"), JSON.stringify({ packages: [".."] }));
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "pi", scope: "packageLocal", action: "remove", actions: {} }],
+      repoRoot: repo,
+    });
+
+    assert.deepEqual(plan.operations.map((op) => op.kind), ["pi-package"]);
+    assert.equal(plan.operations[0].action, "remove");
+    assert.equal(plan.operations[0].status, undefined);
+    assert.equal(plan.reportRows[0].status, "planned");
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("pi package mode update syncs and forces package reinstall plus dependency bootstrap", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-pi-package-update-"));
+  try {
+    const repo = path.join(dir, "repo");
+    await mkdir(path.join(repo, ".pi"), { recursive: true });
+    await writeFile(path.join(repo, ".pi", "settings.json"), JSON.stringify({ packages: [".."] }));
+    await mkdir(path.join(repo, "pi-package", "skills", "atlassian", "scripts", "node_modules"), { recursive: true });
+    await mkdir(path.join(repo, "pi-package", "skills", "web-automation", "scripts", "node_modules", ".bin"), { recursive: true });
+    await writeFile(path.join(repo, "pi-package", "skills", "web-automation", "scripts", "node_modules", ".bin", "cloakbrowser"), "");
+
+    const plan = await buildOperationPlan({
+      selections: [{ clientId: "pi", scope: "packageLocal", action: "update", actions: {} }],
+      repoRoot: repo,
+    });
+
+    assert.equal(plan.operations[0].kind, "sync-pi-package");
+    const packageUpdate = plan.operations.find((op) => op.kind === "pi-package");
+    assert.equal(packageUpdate.action, "update");
+    assert.equal(packageUpdate.status, undefined);
+    assert.deepEqual(
+      plan.reportRows.filter((row) => row.action === "bootstrap-deps").map((row) => [row.item, row.status]).sort(),
+      [["atlassian", "planned"], ["web-automation", "planned"]]
+    );
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});
+
+test("pi package command helper builds exact install and remove argv", () => {
+  assert.deepEqual(piPackageCommand({ action: "install", repoRoot: "/repo", piInstallArg: "" }), ["pi", ["install", "/repo"]]);
+  assert.deepEqual(piPackageCommand({ action: "update", repoRoot: "/repo", piInstallArg: "-l" }), ["pi", ["install", "-l", "/repo"]]);
+  assert.deepEqual(piPackageCommand({ action: "reinstall", repoRoot: "/repo", piInstallArg: "-l" }), ["pi", ["install", "-l", "/repo"]]);
+  assert.deepEqual(piPackageCommand({ action: "remove", repoRoot: "/repo", piInstallArg: "" }), ["pi", ["remove", "/repo"]]);
+  assert.deepEqual(piPackageCommand({ action: "remove", repoRoot: "/repo", piInstallArg: "-l" }), ["pi", ["remove", "-l", "/repo"]]);
+});
+
+test("cli package mode preserves package action and ignores skill narrowing", () => {
+  const output = execFileSync(process.execPath, [
+    path.join(REPO_ROOT, "scripts", "manage-skills.mjs"),
+    "--client", "pi",
+    "--scope", "packageGlobal",
+    "--pi-package",
+    "--skill", "create-plan",
+    "--action", "remove",
+    "--plan-only",
+  ], { cwd: REPO_ROOT, encoding: "utf8" });
+  const plan = JSON.parse(output);
+  assert.deepEqual(plan.operations.map((op) => op.kind), ["pi-package"]);
+  assert.equal(plan.operations[0].action, "remove");
+});
+
+test("cli exits without confirmation when no operations are planned", () => {
+  const output = execFileSync(process.execPath, [
+    path.join(REPO_ROOT, "scripts", "manage-skills.mjs"),
+    "--answers",
+    "/dev/stdin",
+  ], {
+    cwd: REPO_ROOT,
+    encoding: "utf8",
+    input: JSON.stringify({ selections: [{ clientId: "pi", scope: "packageGlobal", action: "skip", actions: {} }] }),
+  });
+  assert.match(output, /No operations planned\./);
+  assert.doesNotMatch(output, /Proceed with these operations/);
+  assert.doesNotMatch(output, /Final report/);
+});
+
+test("validateRemoveTarget rejects paths outside the manifest root", async () => {
+  const dir = await mkdtemp(path.join(tmpdir(), "skill-manager-safe-"));
+  try {
+    await mkdir(path.join(dir, "skills", "create-plan"), { recursive: true });
+    await mkdir(path.join(dir, "outside"), { recursive: true });
+    assert.equal(await validateRemoveTarget(path.join(dir, "skills", "create-plan"), path.join(dir, "skills")), true);
+    await assert.rejects(() => validateRemoveTarget(path.join(dir, "outside"), path.join(dir, "skills")), /outside skills root/);
+    await assert.rejects(() => validateRemoveTarget(dir, dir), /refusing to remove skills root itself/);
+  } finally {
+    await rm(dir, { recursive: true, force: true });
+  }
+});

scripts/verify-reviewer-support.sh

@@ -0,0 +1,186 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)
+export ROOT_DIR
+
+if ! command -v python3 >/dev/null 2>&1; then
+  echo "Missing required command: python3" >&2
+  exit 2
+fi
+
+python3 <<'PY'
+from pathlib import Path
+import os
+import re
+import sys
+
+ROOT = Path(os.environ["ROOT_DIR"])
+WORKFLOWS = {
+    "create-plan": {
+        "payload": "/tmp/plan-",
+        "review_output": "/tmp/plan-review-",
+    },
+    "implement-plan": {
+        "payload": "/tmp/milestone-",
+        "review_output": "/tmp/milestone-review-",
+    },
+    "do-task": {
+        "payload": "/tmp/do-task-",
+        "review_output": "/tmp/do-task-",
+    },
+}
+VARIANTS = ["claude-code", "codex", "cursor", "opencode", "pi"]
+PI_FLAGS = [
+    "--no-session",
+    "--no-skills",
+    "--no-prompt-templates",
+    "--no-extensions",
+    "--no-context-files",
+]
+FORBIDDEN_PI_TOOLS = {"write", "edit", "bash"}
+errors: list[str] = []
+
+
+def compact(text: str) -> str:
+    return re.sub(r"\\\s*\n", " ", text)
+
+
+def has_reviewer_choice(text: str) -> bool:
+    normalized = compact(text)
+    patterns = [
+        r"Reviewer CLI:\s*`codex`,\s*`claude`,\s*`cursor`,\s*`opencode`,\s*`pi`,\s*or\s*`skip`",
+        r"Reviewer CLI\s*\|[^\n]*\bpi\b[^\n]*\bskip\b",
+        r"REVIEWER_CLI[^\n]*`pi`[^\n]*`skip`",
+    ]
+    return any(re.search(pattern, normalized, re.I) for pattern in patterns)
+
+
+def fenced_code_blocks(text: str) -> list[str]:
+    return [match.group(1) for match in re.finditer(r"```(?:bash|sh|shell)?\s*\n(.*?)```", text, re.S)]
+
+
+def pi_command_blocks(text: str) -> list[str]:
+    blocks: list[str] = []
+    for block in fenced_code_blocks(text):
+        normalized = compact(block)
+        if re.search(r"\bpi\s+", normalized) and "--no-session" in normalized:
+            blocks.append(normalized)
+
+    # Fallback for inline examples outside fenced code. Walk from the pi command
+    # until the next blank line/heading instead of relying on a fixed window.
+    lines = text.splitlines()
+    for index, line in enumerate(lines):
+        if re.search(r"\bpi\s+.*--no-session", line):
+            collected = []
+            for candidate in lines[index:]:
+                if collected and (not candidate.strip() or candidate.startswith("#")):
+                    break
+                collected.append(candidate)
+            block = compact("\n".join(collected))
+            if block not in blocks:
+                blocks.append(block)
+    return blocks
+
+
+def block_tools(block: str) -> list[str] | None:
+    match = re.search(r"--tools(?:=|\s+)(['\"]?)([^\s'\"]+)\1", block)
+    if not match:
+        return None
+    return [tool.strip() for tool in match.group(2).split(",") if tool.strip()]
+
+
+def block_has_exact_tools(block: str) -> bool:
+    return block_tools(block) == ["read", "grep", "find", "ls"]
+
+
+for workflow, spec in WORKFLOWS.items():
+    for variant in VARIANTS:
+        path = ROOT / "skills" / workflow / variant / "SKILL.md"
+        if not path.exists():
+            errors.append(f"missing workflow variant: {path}")
+            continue
+        text = path.read_text()
+
+        if not has_reviewer_choice(text):
+            errors.append(f"{path}: reviewer choices must include pi and skip")
+
+        if "pi/<pi-model-name>" not in text and "pi/claude-opus-4-7" not in text:
+            errors.append(f"{path}: must document pi/<pi-model-name> reviewer shorthand")
+
+        if "pi --list-models [search]" not in text:
+            errors.append(f"{path}: must mention pi --list-models [search] for unavailable models")
+
+        if not re.search(r"reviewer model is configured independently|model is configured independently", text, re.I):
+            errors.append(f"{path}: must state Pi reviewer model is configured independently")
+
+        blocks = pi_command_blocks(text)
+        if not blocks:
+            errors.append(f"{path}: missing isolated pi reviewer command block")
+            continue
+
+        matching_blocks = [block for block in blocks if spec["payload"] in block]
+        if not matching_blocks:
+            errors.append(f"{path}: pi reviewer command must reference {spec['payload']} payload path")
+            matching_blocks = blocks
+
+        if not any(all(flag in block for flag in PI_FLAGS) for block in matching_blocks):
+            errors.append(f"{path}: pi reviewer command missing one or more isolation flags")
+
+        if not any(block_has_exact_tools(block) for block in matching_blocks):
+            errors.append(f"{path}: pi reviewer command must use exactly --tools read,grep,find,ls")
+
+        # The forbidden-tool check is scoped to the --tools allowlist in the Pi
+        # reviewer command. Prose that says these tools are forbidden is allowed.
+        for block in matching_blocks:
+            tools = block_tools(block)
+            if tools is None:
+                continue
+            forbidden = sorted(set(tools) & FORBIDDEN_PI_TOOLS)
+            if forbidden:
+                errors.append(f"{path}: pi reviewer command includes forbidden tools: {', '.join(forbidden)}")
+
+        if variant != "pi" and ".pi/skills/reviewer-runtime/pi" in text:
+            errors.append(f"{path}: non-Pi variant must not use Pi reviewer-runtime helper path as its own runtime")
+
+for variant in VARIANTS:
+    template = ROOT / "skills" / "do-task" / variant / "templates" / "task-plan.md"
+    if not template.exists():
+        errors.append(f"missing do-task template: {template}")
+        continue
+    text = template.read_text()
+    if "Reviewer CLI | codex \\| claude \\| cursor \\| opencode \\| pi" not in text:
+        errors.append(f"{template}: Reviewer CLI metadata must include pi")
+
+canonical = ROOT / "docs" / "PI-COMMON-REVIEWER.md"
+if not canonical.exists():
+    errors.append("docs/PI-COMMON-REVIEWER.md is missing")
+else:
+    text = canonical.read_text()
+    for flag in PI_FLAGS:
+        if flag not in text:
+            errors.append(f"docs/PI-COMMON-REVIEWER.md: missing {flag}")
+    if "--tools read,grep,find,ls" not in compact(text):
+        errors.append("docs/PI-COMMON-REVIEWER.md: missing exact read-only tool allowlist")
+    if "MUST NOT include `write`, `edit`, or `bash`" not in text:
+        errors.append("docs/PI-COMMON-REVIEWER.md: must forbid write/edit/bash tools")
+
+for doc in ["CREATE-PLAN.md", "IMPLEMENT-PLAN.md", "DO-TASK.md"]:
+    path = ROOT / "docs" / doc
+    if not path.exists():
+        errors.append(f"docs/{doc} is missing")
+        continue
+    text = path.read_text()
+    if "PI-COMMON-REVIEWER.md" not in text:
+        errors.append(f"docs/{doc}: must link to PI-COMMON-REVIEWER.md")
+    if "pi/claude-opus-4-7" not in text and "pi/<pi-model-name>" not in text:
+        errors.append(f"docs/{doc}: must document Pi reviewer shorthand")
+
+if errors:
+    print("Reviewer support verification failed:", file=sys.stderr)
+    for error in errors:
+        print(f"- {error}", file=sys.stderr)
+    sys.exit(1)
+
+print("reviewer support verified")
+PY

skills/create-plan/claude-code/SKILL.md

@@ -33,6 +33,19 @@ If any dependency is missing, stop immediately and return:
 
 ### Phase 3: Configure Reviewer
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
+
 If the user has already specified a reviewer CLI and model (e.g., "create a plan, review with codex o4-mini"), use those values. Otherwise, ask:
 
 1. **Which CLI should review the plan?**
@@ -152,6 +165,18 @@ Write the reviewer invocation to `/tmp/plan-review-${REVIEW_ID}.sh` as a bash sc
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/plan-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 ```bash
@@ -281,7 +306,7 @@ jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_I
 ```
 
 - If `REVIEWER_CLI=codex`, extract `CODEX_SESSION_ID` from `/tmp/plan-review-${REVIEW_ID}.runner.out` after the helper or fallback run. If the review text is only in `.runner.out`, move or copy the actual review body into `/tmp/plan-review-${REVIEW_ID}.md` before verdict parsing.
-- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
+- If `REVIEWER_CLI=claude` or `REVIEWER_CLI=pi`, promote stdout captured by the helper or fallback runner into the markdown review file:
 
 ```bash
 cp /tmp/plan-review-${REVIEW_ID}.runner.out /tmp/plan-review-${REVIEW_ID}.md
@@ -331,6 +356,18 @@ If a revision contradicts the user's explicit requirements, skip it and note it
 
 Rewrite `/tmp/plan-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call with prior-round context (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "You previously reviewed this plan and requested revisions. Read the updated payload at /tmp/plan-${REVIEW_ID}.md and re-review using the same ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Resume the existing session:

skills/create-plan/codex/SKILL.md

@@ -58,6 +58,19 @@ If any dependency is missing, stop and return:
 
 ### Phase 3: Configure Reviewer
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
+
 If the user has already specified a reviewer CLI and model (e.g., "create a plan, review with claude sonnet"), use those values. Otherwise, ask:
 
 1. **Which CLI should review the plan?**
@@ -175,6 +188,18 @@ Write the reviewer invocation to `/tmp/plan-review-${REVIEW_ID}.sh` as a bash sc
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/plan-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 ```bash
@@ -304,7 +329,7 @@ jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_I
 ```
 
 - If `REVIEWER_CLI=codex`, extract `CODEX_SESSION_ID` from `/tmp/plan-review-${REVIEW_ID}.runner.out` after the helper or fallback run. If the review text is only in `.runner.out`, move or copy the actual review body into `/tmp/plan-review-${REVIEW_ID}.md` before verdict parsing.
-- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
+- If `REVIEWER_CLI=claude` or `REVIEWER_CLI=pi`, promote stdout captured by the helper or fallback runner into the markdown review file:
 
 ```bash
 cp /tmp/plan-review-${REVIEW_ID}.runner.out /tmp/plan-review-${REVIEW_ID}.md
@@ -354,6 +379,18 @@ If a revision contradicts the user's explicit requirements, skip it and note it
 
 Rewrite `/tmp/plan-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call with prior-round context (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "You previously reviewed this plan and requested revisions. Read the updated payload at /tmp/plan-${REVIEW_ID}.md and re-review using the same ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Resume the existing session:

skills/create-plan/cursor/SKILL.md

@@ -15,7 +15,7 @@ This skill wraps the current Superpowers flow for the Cursor Agent CLI (`cursor-
 3. Review the plan iteratively with a second model/provider
 4. Persist a local execution package in `ai_plan/YYYY-MM-DD-<short-title>/`
 
-**Core principle:** Cursor Agent CLI discovers skills from `.cursor/skills/` (repo-local or `~/.cursor/skills/` global). It also reads `AGENTS.md` at the repo root for additional instructions.
+**Core principle:** Cursor Agent CLI discovers skills from `.cursor/skills/` (repo-local), `~/.cursor/skills/` (global), and installed Cursor plugin cache entries. It also reads `AGENTS.md` at the repo root for additional instructions.
 
 ## Prerequisite Check (MANDATORY)
 
@@ -23,7 +23,7 @@ Required:
 - Cursor Agent CLI: `cursor-agent --version` (install via `curl https://cursor.com/install -fsS | bash`). The binary is `cursor-agent` (installed to `~/.local/bin/`). Some environments alias it as `cursor agent` (subcommand of the Cursor IDE CLI) — both forms work, but this skill uses `cursor-agent` throughout.
 - `jq` (required only if using `cursor` as the reviewer CLI): `jq --version` (install via `brew install jq` or your package manager)
 - Superpowers repo: `https://github.com/obra/superpowers`
-- Superpowers skills installed under `.cursor/skills/` (repo-local) or `~/.cursor/skills/` (global)
+- Superpowers skills available from the Cursor plugin cache, `.cursor/skills/` (repo-local), or `~/.cursor/skills/` (global). Do not install both the plugin and a manual Superpowers copy, or Cursor may show duplicate skill entries.
 - `superpowers:brainstorming`
 - `superpowers:writing-plans`
 
@@ -31,15 +31,15 @@ Verify before proceeding:
 
 ```bash
 cursor-agent --version
-test -f .cursor/skills/superpowers/skills/brainstorming/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/brainstorming/SKILL.md
-test -f .cursor/skills/superpowers/skills/writing-plans/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/writing-plans/SKILL.md
+test -f .cursor/skills/superpowers/skills/brainstorming/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/brainstorming/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/brainstorming/SKILL.md' -print -quit 2>/dev/null | grep -q .
+test -f .cursor/skills/superpowers/skills/writing-plans/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/writing-plans/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/writing-plans/SKILL.md' -print -quit 2>/dev/null | grep -q .
 # Only if using cursor as reviewer CLI:
 # jq --version
 ```
 
 If any dependency is missing, stop and return:
 
-`Missing dependency: Superpowers planning skills are required (superpowers:brainstorming, superpowers:writing-plans). Install from https://github.com/obra/superpowers and copy into .cursor/skills/ or ~/.cursor/skills/, then retry.`
+`Missing dependency: Superpowers planning skills are required (superpowers:brainstorming, superpowers:writing-plans). Install the Cursor Superpowers plugin or install Superpowers under .cursor/skills/ or ~/.cursor/skills/, then retry.`
 
 ## Required Skill Invocation Rules
 
@@ -59,6 +59,19 @@ If any dependency is missing, stop and return:
 
 ### Phase 3: Configure Reviewer
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
+
 If the user has already specified a reviewer CLI and model (e.g., "create a plan, review with codex o4-mini"), use those values. Otherwise, ask:
 
 1. **Which CLI should review the plan?**
@@ -181,6 +194,18 @@ Write the reviewer invocation to `/tmp/plan-review-${REVIEW_ID}.sh` as a bash sc
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/plan-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 ```bash
@@ -316,7 +341,7 @@ jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_I
 ```
 
 - If `REVIEWER_CLI=codex`, extract `CODEX_SESSION_ID` from `/tmp/plan-review-${REVIEW_ID}.runner.out` after the helper or fallback run. If the review text is only in `.runner.out`, move or copy the actual review body into `/tmp/plan-review-${REVIEW_ID}.md` before verdict parsing.
-- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
+- If `REVIEWER_CLI=claude` or `REVIEWER_CLI=pi`, promote stdout captured by the helper or fallback runner into the markdown review file:
 
 ```bash
 cp /tmp/plan-review-${REVIEW_ID}.runner.out /tmp/plan-review-${REVIEW_ID}.md
@@ -364,6 +389,18 @@ If a revision contradicts the user's explicit requirements, skip it and note it
 
 Rewrite `/tmp/plan-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call with prior-round context (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "You previously reviewed this plan and requested revisions. Read the updated payload at /tmp/plan-${REVIEW_ID}.md and re-review using the same ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Resume the existing session:

skills/create-plan/opencode/SKILL.md

@@ -9,18 +9,19 @@ Create and maintain a local plan folder under `ai_plan/` at project root.
 
 ## Prerequisite Check (MANDATORY)
 
-This OpenCode variant depends on Superpowers skills being installed via OpenCode's native skill system.
+This OpenCode variant depends on Superpowers skills being available through OpenCode's native skill system.
 
 Required:
 - Superpowers repo: `https://github.com/obra/superpowers`
-- OpenCode Superpowers skills symlink: `~/.config/opencode/skills/superpowers`
+- OpenCode Superpowers skills available at `~/.agents/skills/superpowers` or `~/.config/opencode/skills/superpowers`
 - `superpowers/brainstorming`
 - `superpowers/writing-plans`
 
 Verify before proceeding:
 
 ```bash
-ls -l ~/.config/opencode/skills/superpowers
+test -f ~/.agents/skills/superpowers/brainstorming/SKILL.md || test -f ~/.config/opencode/skills/superpowers/brainstorming/SKILL.md
+test -f ~/.agents/skills/superpowers/writing-plans/SKILL.md || test -f ~/.config/opencode/skills/superpowers/writing-plans/SKILL.md
 ```
 
 If dependencies are missing, stop immediately and return:
@@ -64,6 +65,19 @@ If the user has already specified a reviewer CLI and model (e.g., "create a plan
 
 Store the chosen `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phase 7 (Iterative Plan Review).
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
+
 ### Phase 5: Design (REQUIRED SUB-SKILL)
 
 Use OpenCode's native skill tool to load:
@@ -169,6 +183,18 @@ Write the reviewer invocation to `/tmp/plan-review-${REVIEW_ID}.sh` as a bash sc
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/plan-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 ```bash
@@ -298,7 +324,7 @@ jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_I
 ```
 
 - If `REVIEWER_CLI=codex`, extract `CODEX_SESSION_ID` from `/tmp/plan-review-${REVIEW_ID}.runner.out` after the helper or fallback run. If the review text is only in `.runner.out`, move or copy the actual review body into `/tmp/plan-review-${REVIEW_ID}.md` before verdict parsing.
-- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
+- If `REVIEWER_CLI=claude` or `REVIEWER_CLI=pi`, promote stdout captured by the helper or fallback runner into the markdown review file:
 
 ```bash
 cp /tmp/plan-review-${REVIEW_ID}.runner.out /tmp/plan-review-${REVIEW_ID}.md
@@ -346,6 +372,18 @@ If a revision contradicts the user's explicit requirements, skip it and note it
 
 Rewrite `/tmp/plan-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call with prior-round context (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "You previously reviewed this plan and requested revisions. Read the updated payload at /tmp/plan-${REVIEW_ID}.md and re-review using the same ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Resume the existing session:

skills/create-plan/pi/SKILL.md

@@ -85,6 +85,14 @@ Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for the review loop.
 
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
 When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the pi model running this workflow. Use any configured pi model string, including provider-qualified model IDs. If the reviewer model or provider is unavailable, surface the review helper stderr/status and ask for a configured model; use `pi --list-models [search]` to inspect configured models.
 
 The pi reviewer command rendered into `/tmp/plan-review-${REVIEW_ID}.sh` must be isolated and read-only:

skills/do-task/claude-code/SKILL.md

@@ -111,6 +111,19 @@ If the user has already specified a reviewer CLI and model (e.g., "do task X, re
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phases 5 and 8.
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ### Phase 4: Initialize Plan Workspace
 
 **PLAN MODE CHECK:** If currently in plan mode:
@@ -344,7 +357,7 @@ This subroutine is invoked twice per `do-task` run: once in Phase 5 (`REVIEW_KIN
 | `REVIEW_ID` | 8-char hex (from `uuidgen`); reused across rounds of the same loop |
 | `PAYLOAD_PATH` | `/tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md` |
 | `PROMPT_TEMPLATE` | `PLAN_REVIEW_PROMPT` or `IMPL_REVIEW_PROMPT` |
-| `REVIEWER_CLI` | `codex` \| `claude` \| `cursor` \| `opencode` |
+| `REVIEWER_CLI` | `codex` \| `claude` \| `cursor` \| `opencode` \| `pi` |
 | `REVIEWER_MODEL` | Model name |
 | `MAX_ROUNDS` | Default 10 |
 | `SESSION_ID_VAR` | `CODEX_PLAN_SESSION_ID` \| `CODEX_IMPL_SESSION_ID` \| `CURSOR_PLAN_SESSION_ID` \| `CURSOR_IMPL_SESSION_ID` \| `OPENCODE_PLAN_SESSION_ID` \| `OPENCODE_IMPL_SESSION_ID` |
@@ -437,6 +450,18 @@ Write the reviewer invocation to `/tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Round 1 — fresh `codex exec`:
@@ -633,7 +658,7 @@ After the command completes:
   cp /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.runner.out \
      /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.md
   ```
-- `claude`: promote `.runner.out` into the `.md` file:
+- `claude` or `pi`: promote `.runner.out` into the `.md` file:
   ```bash
   cp /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.runner.out \
      /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.md

skills/do-task/claude-code/templates/task-plan.md

@@ -9,7 +9,7 @@
 | Created | YYYY-MM-DD |
 | Slug | YYYY-MM-DD-<slug> |
 | Runtime | claude-code |
-| Reviewer CLI | codex \| claude \| cursor \| opencode |
+| Reviewer CLI | codex \| claude \| cursor \| opencode \| pi |
 | Reviewer Model | <model> |
 | MAX_ROUNDS | 10 |
 | Branch Strategy | current-branch \| worktree |

skills/do-task/codex/SKILL.md

@@ -133,6 +133,19 @@ If the user has already specified a reviewer CLI and model (e.g., "do task X, re
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phases 5 and 8.
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ### Phase 4: Initialize Plan Workspace
 
 Codex has no plan-mode concept; there is no plan-mode guard here.
@@ -363,7 +376,7 @@ This subroutine is invoked twice per `do-task` run: once in Phase 5 (`REVIEW_KIN
 | `REVIEW_ID` | 8-char hex (from `uuidgen`); reused across rounds of the same loop |
 | `PAYLOAD_PATH` | `/tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md` |
 | `PROMPT_TEMPLATE` | `PLAN_REVIEW_PROMPT` or `IMPL_REVIEW_PROMPT` |
-| `REVIEWER_CLI` | `codex` \| `claude` \| `cursor` \| `opencode` |
+| `REVIEWER_CLI` | `codex` \| `claude` \| `cursor` \| `opencode` \| `pi` |
 | `REVIEWER_MODEL` | Model name |
 | `MAX_ROUNDS` | Default 10 |
 | `SESSION_ID_VAR` | `CODEX_PLAN_SESSION_ID` \| `CODEX_IMPL_SESSION_ID` \| `CURSOR_PLAN_SESSION_ID` \| `CURSOR_IMPL_SESSION_ID` \| `OPENCODE_PLAN_SESSION_ID` \| `OPENCODE_IMPL_SESSION_ID` |
@@ -456,6 +469,18 @@ Write the reviewer invocation to `/tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Round 1 — fresh `codex exec`:
@@ -652,7 +677,7 @@ After the command completes:
   cp /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.runner.out \
      /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.md
   ```
-- `claude`: promote `.runner.out` into the `.md` file:
+- `claude` or `pi`: promote `.runner.out` into the `.md` file:
   ```bash
   cp /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.runner.out \
      /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.md

skills/do-task/codex/templates/task-plan.md

@@ -9,7 +9,7 @@
 | Created | YYYY-MM-DD |
 | Slug | YYYY-MM-DD-<slug> |
 | Runtime | codex |
-| Reviewer CLI | codex \| claude \| cursor \| opencode |
+| Reviewer CLI | codex \| claude \| cursor \| opencode \| pi |
 | Reviewer Model | <model> |
 | MAX_ROUNDS | 10 |
 | Branch Strategy | current-branch \| worktree |

skills/do-task/cursor/SKILL.md

@@ -9,7 +9,7 @@ Execute an ad-hoc user prompt end-to-end: parse → clarify → plan (with revie
 
 This is a single-artifact sibling of `create-plan` + `implement-plan`. Unlike `implement-plan`, `do-task` operates on one persistent `task-plan.md` (not a full milestone plan) and defaults to the **current branch** (not a worktree).
 
-**Core principle:** Cursor Agent CLI discovers skills from `.cursor/skills/` (repo-local) or `~/.cursor/skills/` (global). It also reads `AGENTS.md` at the repo root for additional instructions.
+**Core principle:** Cursor Agent CLI discovers skills from `.cursor/skills/` (repo-local), `~/.cursor/skills/` (global), and installed Cursor plugin cache entries. It also reads `AGENTS.md` at the repo root for additional instructions.
 
 ## Prerequisite Check (MANDATORY)
 
@@ -17,7 +17,7 @@ Required:
 - Cursor Agent CLI: `cursor-agent --version` (install via `curl https://cursor.com/install -fsS | bash`). Binary is `cursor-agent`; the alias `cursor agent` also works.
 - `jq` (**required** — `do-task` always parses JSON output from at least the cursor reviewer branch, and other reviewers may produce JSON). Install via `brew install jq` (macOS) or your package manager. Verify: `jq --version`.
 - Superpowers repo: `https://github.com/obra/superpowers`
-- Superpowers skills installed under `.cursor/skills/` (repo-local) or `~/.cursor/skills/` (global)
+- Superpowers skills available from the Cursor plugin cache, `.cursor/skills/` (repo-local), or `~/.cursor/skills/` (global). Do not install both the plugin and a manual Superpowers copy, or Cursor may show duplicate skill entries.
 - `superpowers:brainstorming`
 - `superpowers:test-driven-development`
 - `superpowers:verification-before-completion`
@@ -31,15 +31,15 @@ Verify before proceeding:
 ```bash
 cursor-agent --version
 jq --version
-test -f .cursor/skills/superpowers/skills/brainstorming/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/brainstorming/SKILL.md
-test -f .cursor/skills/superpowers/skills/test-driven-development/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/test-driven-development/SKILL.md
-test -f .cursor/skills/superpowers/skills/verification-before-completion/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/verification-before-completion/SKILL.md
-test -f .cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md
+test -f .cursor/skills/superpowers/skills/brainstorming/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/brainstorming/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/brainstorming/SKILL.md' -print -quit 2>/dev/null | grep -q .
+test -f .cursor/skills/superpowers/skills/test-driven-development/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/test-driven-development/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/test-driven-development/SKILL.md' -print -quit 2>/dev/null | grep -q .
+test -f .cursor/skills/superpowers/skills/verification-before-completion/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/verification-before-completion/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/verification-before-completion/SKILL.md' -print -quit 2>/dev/null | grep -q .
+test -f .cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/finishing-a-development-branch/SKILL.md' -print -quit 2>/dev/null | grep -q .
 ```
 
 If any required dependency is missing, stop immediately and return:
 
-`Missing dependency: [specific missing item]. Install Cursor Agent CLI, jq, and Superpowers skills under .cursor/skills/ or ~/.cursor/skills/, then retry.`
+`Missing dependency: [specific missing item]. Install Cursor Agent CLI, jq, and the Cursor Superpowers plugin or Superpowers skills under .cursor/skills/ or ~/.cursor/skills/, then retry.`
 
 ## Required Skill Invocation Rules
 
@@ -132,6 +132,19 @@ If the user has already specified a reviewer CLI and model (e.g., "do task X, re
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phases 5 and 8.
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ### Phase 4: Initialize Plan Workspace
 
 Cursor Agent CLI has no plan-mode concept; there is no plan-mode guard here.
@@ -366,7 +379,7 @@ This subroutine is invoked twice per `do-task` run: once in Phase 5 (`REVIEW_KIN
 | `REVIEW_ID` | 8-char hex (from `uuidgen`); reused across rounds of the same loop |
 | `PAYLOAD_PATH` | `/tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md` |
 | `PROMPT_TEMPLATE` | `PLAN_REVIEW_PROMPT` or `IMPL_REVIEW_PROMPT` |
-| `REVIEWER_CLI` | `codex` \| `claude` \| `cursor` \| `opencode` |
+| `REVIEWER_CLI` | `codex` \| `claude` \| `cursor` \| `opencode` \| `pi` |
 | `REVIEWER_MODEL` | Model name |
 | `MAX_ROUNDS` | Default 10 |
 | `SESSION_ID_VAR` | `CODEX_PLAN_SESSION_ID` \| `CODEX_IMPL_SESSION_ID` \| `CURSOR_PLAN_SESSION_ID` \| `CURSOR_IMPL_SESSION_ID` \| `OPENCODE_PLAN_SESSION_ID` \| `OPENCODE_IMPL_SESSION_ID` |
@@ -463,6 +476,18 @@ Write the reviewer invocation to `/tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Round 1 — fresh `codex exec`:
@@ -659,7 +684,7 @@ After the command completes:
   cp /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.runner.out \
      /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.md
   ```
-- `claude`: promote `.runner.out` into the `.md` file:
+- `claude` or `pi`: promote `.runner.out` into the `.md` file:
   ```bash
   cp /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.runner.out \
      /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.md

skills/do-task/cursor/templates/task-plan.md

@@ -9,7 +9,7 @@
 | Created | YYYY-MM-DD |
 | Slug | YYYY-MM-DD-<slug> |
 | Runtime | cursor |
-| Reviewer CLI | codex \| claude \| cursor \| opencode |
+| Reviewer CLI | codex \| claude \| cursor \| opencode \| pi |
 | Reviewer Model | <model> |
 | MAX_ROUNDS | 10 |
 | Branch Strategy | current-branch \| worktree |

skills/do-task/opencode/SKILL.md

@@ -9,14 +9,14 @@ Execute an ad-hoc user prompt end-to-end: parse → clarify → plan (with revie
 
 This is a single-artifact sibling of `create-plan` + `implement-plan`. Unlike `implement-plan`, `do-task` operates on one persistent `task-plan.md` (not a full milestone plan) and defaults to the **current branch** (not a worktree).
 
-**Core principle:** OpenCode loads skills through its native skill tool from `~/.config/opencode/skills/`. Sub-skill invocations use OpenCode's native mechanism — not Claude's `Skill` tool, not Codex's native-discovery patterns, not Cursor's workspace discovery.
+**Core principle:** OpenCode loads skills through its native skill tool. Local skills live under `~/.config/opencode/skills/`, and OpenCode can also expose shared agent skills from `~/.agents/skills/`. Sub-skill invocations use OpenCode's native mechanism — not Claude's `Skill` tool, not Cursor's workspace discovery.
 
 ## Prerequisite Check (MANDATORY)
 
 Required:
 - OpenCode CLI: `opencode --version` (install via your package manager or `brew install opencode`).
 - Superpowers repo: `https://github.com/obra/superpowers`
-- OpenCode Superpowers skills symlink: `~/.config/opencode/skills/superpowers`
+- OpenCode Superpowers skills available at `~/.agents/skills/superpowers` or `~/.config/opencode/skills/superpowers`
 - `superpowers/brainstorming`
 - `superpowers/test-driven-development`
 - `superpowers/verification-before-completion`
@@ -29,11 +29,10 @@ Verify before proceeding:
 
 ```bash
 opencode --version
-ls -l ~/.config/opencode/skills/superpowers
-test -f ~/.config/opencode/skills/superpowers/brainstorming/SKILL.md
-test -f ~/.config/opencode/skills/superpowers/test-driven-development/SKILL.md
-test -f ~/.config/opencode/skills/superpowers/verification-before-completion/SKILL.md
-test -f ~/.config/opencode/skills/superpowers/finishing-a-development-branch/SKILL.md
+test -f ~/.agents/skills/superpowers/brainstorming/SKILL.md || test -f ~/.config/opencode/skills/superpowers/brainstorming/SKILL.md
+test -f ~/.agents/skills/superpowers/test-driven-development/SKILL.md || test -f ~/.config/opencode/skills/superpowers/test-driven-development/SKILL.md
+test -f ~/.agents/skills/superpowers/verification-before-completion/SKILL.md || test -f ~/.config/opencode/skills/superpowers/verification-before-completion/SKILL.md
+test -f ~/.agents/skills/superpowers/finishing-a-development-branch/SKILL.md || test -f ~/.config/opencode/skills/superpowers/finishing-a-development-branch/SKILL.md
 ```
 
 If any required dependency is missing, stop immediately and return:
@@ -46,7 +45,7 @@ If any required dependency is missing, stop immediately and return:
 - Announce skill usage explicitly:
   - `I've read the [Skill Name] skill and I'm using it to [purpose].`
 - For skills with checklists, track checklist items explicitly in conversation.
-- Do NOT use Claude's `Skill` tool syntax, Codex's `~/.agents/skills/` native-discovery paths, or Cursor's workspace discovery. OpenCode's skill system is independent.
+- Do NOT use Claude's `Skill` tool syntax or Cursor's workspace discovery. OpenCode's skill system may expose shared files from `~/.agents/skills/`, but invocation still goes through OpenCode's native skill mechanism.
 
 ## Trigger Phrase Detection
 
@@ -132,6 +131,19 @@ If the user has already specified a reviewer CLI and model (e.g., "do task X, re
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phases 5 and 8.
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ### Phase 4: Initialize Plan Workspace
 
 OpenCode has no plan-mode concept; there is no plan-mode guard here.
@@ -362,7 +374,7 @@ This subroutine is invoked twice per `do-task` run: once in Phase 5 (`REVIEW_KIN
 | `REVIEW_ID` | 8-char hex (from `uuidgen`); reused across rounds of the same loop |
 | `PAYLOAD_PATH` | `/tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md` |
 | `PROMPT_TEMPLATE` | `PLAN_REVIEW_PROMPT` or `IMPL_REVIEW_PROMPT` |
-| `REVIEWER_CLI` | `codex` \| `claude` \| `cursor` \| `opencode` |
+| `REVIEWER_CLI` | `codex` \| `claude` \| `cursor` \| `opencode` \| `pi` |
 | `REVIEWER_MODEL` | Model name |
 | `MAX_ROUNDS` | Default 10 |
 | `SESSION_ID_VAR` | `CODEX_PLAN_SESSION_ID` \| `CODEX_IMPL_SESSION_ID` \| `CURSOR_PLAN_SESSION_ID` \| `CURSOR_IMPL_SESSION_ID` \| `OPENCODE_PLAN_SESSION_ID` \| `OPENCODE_IMPL_SESSION_ID` |
@@ -455,6 +467,18 @@ Write the reviewer invocation to `/tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/do-task-${REVIEW_KIND}-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Round 1 — fresh `codex exec`:
@@ -651,7 +675,7 @@ After the command completes:
   cp /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.runner.out \
      /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.md
   ```
-- `claude`: promote `.runner.out` into the `.md` file:
+- `claude` or `pi`: promote `.runner.out` into the `.md` file:
   ```bash
   cp /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.runner.out \
      /tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.md
@@ -770,7 +794,7 @@ Review History is append-only.
 
 ## Variant Hardening Notes — OpenCode
 
-- Must use OpenCode's native skill tool for sub-skill invocation. Do NOT use Claude's `Skill` tool syntax or Codex's `~/.agents/skills/` paths.
+- Must use OpenCode's native skill tool for sub-skill invocation. Do NOT use Claude's `Skill` tool syntax. OpenCode may load shared skill files from `~/.agents/skills/`, but invocation is still OpenCode-native.
 - Phase 1 includes a Bootstrap Superpowers Context step that lists installed skills and confirms `superpowers/brainstorming`, `superpowers/test-driven-development`, `superpowers/verification-before-completion`, and `superpowers/finishing-a-development-branch` are discoverable before any other phase runs.
 - Helper paths are `~/.config/opencode/skills/reviewer-runtime/{run-review.sh,notify-telegram.sh}`.
 - OpenCode reviewer CLI branch (when `REVIEWER_CLI=opencode`):
@@ -784,7 +808,7 @@ Review History is append-only.
 ## Common Mistakes
 
 - Skipping the Bootstrap Superpowers Context step in Phase 1 (breaks native skill discovery).
-- Using Claude `Skill` tool syntax or Codex `~/.agents/skills/` paths.
+- Using Claude `Skill` tool syntax, or treating shared `~/.agents/skills/` files as anything other than OpenCode-native skill entries.
 - Forgetting to set `--agent plan` on opencode reviewer calls (would use the default `build` agent which can write files).
 - Asking multiple clarifying questions in a single message.
 - Skipping the per-payload secret scan because "the previous round was clean".

skills/do-task/opencode/templates/task-plan.md

@@ -9,7 +9,7 @@
 | Created | YYYY-MM-DD |
 | Slug | YYYY-MM-DD-<slug> |
 | Runtime | opencode |
-| Reviewer CLI | codex \| claude \| cursor \| opencode |
+| Reviewer CLI | codex \| claude \| cursor \| opencode \| pi |
 | Reviewer Model | <model> |
 | MAX_ROUNDS | 10 |
 | Branch Strategy | current-branch \| worktree |

skills/do-task/pi/SKILL.md

@@ -103,6 +103,14 @@ Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS`.
 
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
 When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the pi model running this workflow. Use any configured pi model string, including provider-qualified model IDs. If the reviewer model or provider is unavailable, surface the review helper stderr/status and ask for a configured model; use `pi --list-models [search]` to inspect configured models.
 
 The pi reviewer command rendered into `/tmp/do-task-${REVIEW_KIND}-review-${REVIEW_ID}.sh` must be isolated and read-only:

skills/implement-plan/claude-code/SKILL.md

@@ -61,6 +61,19 @@ If the user has already specified a reviewer CLI and model (e.g., "implement the
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS`. These values are fixed for the entire run.
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ### Phase 3: Set Up Worktree (REQUIRED SUB-SKILL)
 
 Invoke `superpowers:using-git-worktrees` explicitly.
@@ -235,6 +248,18 @@ Write the reviewer invocation to `/tmp/milestone-review-${REVIEW_ID}.sh` as a ba
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/milestone-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 ```bash
@@ -372,7 +397,7 @@ jq -r '.result' /tmp/milestone-review-${REVIEW_ID}.json > /tmp/milestone-review-
 ```
 
 - If `REVIEWER_CLI=codex`, extract `CODEX_SESSION_ID` from `/tmp/milestone-review-${REVIEW_ID}.runner.out` after the helper or fallback run. If the review text is only in `.runner.out`, move or copy the actual review body into `/tmp/milestone-review-${REVIEW_ID}.md` before verdict parsing.
-- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
+- If `REVIEWER_CLI=claude` or `REVIEWER_CLI=pi`, promote stdout captured by the helper or fallback runner into the markdown review file:
 
 ```bash
 cp /tmp/milestone-review-${REVIEW_ID}.runner.out /tmp/milestone-review-${REVIEW_ID}.md
@@ -424,6 +449,18 @@ If a revision contradicts the user's explicit requirements, skip it and note it
 
 Rewrite `/tmp/milestone-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call with prior-round context (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "You previously reviewed this milestone and requested revisions. Read the updated payload at /tmp/milestone-${REVIEW_ID}.md and re-review using the same ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Resume the existing session:

skills/implement-plan/codex/SKILL.md

@@ -94,6 +94,19 @@ If the user has already specified a reviewer CLI and model (e.g., "implement the
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS`. These values are fixed for the entire run.
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ### Phase 3: Set Up Worktree (REQUIRED SUB-SKILL)
 
 Invoke `superpowers:using-git-worktrees`.
@@ -268,6 +281,18 @@ Write the reviewer invocation to `/tmp/milestone-review-${REVIEW_ID}.sh` as a ba
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/milestone-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 ```bash
@@ -405,7 +430,7 @@ jq -r '.result' /tmp/milestone-review-${REVIEW_ID}.json > /tmp/milestone-review-
 ```
 
 - If `REVIEWER_CLI=codex`, extract `CODEX_SESSION_ID` from `/tmp/milestone-review-${REVIEW_ID}.runner.out` after the helper or fallback run. If the review text is only in `.runner.out`, move or copy the actual review body into `/tmp/milestone-review-${REVIEW_ID}.md` before verdict parsing.
-- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
+- If `REVIEWER_CLI=claude` or `REVIEWER_CLI=pi`, promote stdout captured by the helper or fallback runner into the markdown review file:
 
 ```bash
 cp /tmp/milestone-review-${REVIEW_ID}.runner.out /tmp/milestone-review-${REVIEW_ID}.md
@@ -457,6 +482,18 @@ If a revision contradicts the user's explicit requirements, skip it and note it
 
 Rewrite `/tmp/milestone-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call with prior-round context (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "You previously reviewed this milestone and requested revisions. Read the updated payload at /tmp/milestone-${REVIEW_ID}.md and re-review using the same ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Resume the existing session:

skills/implement-plan/cursor/SKILL.md

@@ -16,7 +16,7 @@ This skill wraps the Superpowers execution flow for the Cursor Agent CLI (`curso
 4. Review each milestone with a second model/provider
 5. Commit approved milestones, merge to parent branch, and delete worktree
 
-**Core principle:** Cursor Agent CLI discovers skills from `.cursor/skills/` (repo-local or `~/.cursor/skills/` global). It also reads `AGENTS.md` at the repo root for additional instructions.
+**Core principle:** Cursor Agent CLI discovers skills from `.cursor/skills/` (repo-local), `~/.cursor/skills/` (global), and installed Cursor plugin cache entries. It also reads `AGENTS.md` at the repo root for additional instructions.
 
 ## Prerequisite Check (MANDATORY)
 
@@ -28,7 +28,7 @@ Required:
 - `milestone-plan.md` exists in plan folder
 - `story-tracker.md` exists in plan folder
 - Git repo with worktree support: `git worktree list`
-- Superpowers skills installed under `.cursor/skills/` (repo-local) or `~/.cursor/skills/` (global)
+- Superpowers skills available from the Cursor plugin cache, `.cursor/skills/` (repo-local), or `~/.cursor/skills/` (global). Do not install both the plugin and a manual Superpowers copy, or Cursor may show duplicate skill entries.
 - Superpowers execution skills:
   - `superpowers:executing-plans`
   - `superpowers:using-git-worktrees`
@@ -39,17 +39,17 @@ Verify before proceeding:
 
 ```bash
 cursor-agent --version
-test -f .cursor/skills/superpowers/skills/executing-plans/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/executing-plans/SKILL.md
-test -f .cursor/skills/superpowers/skills/using-git-worktrees/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/using-git-worktrees/SKILL.md
-test -f .cursor/skills/superpowers/skills/verification-before-completion/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/verification-before-completion/SKILL.md
-test -f .cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md
+test -f .cursor/skills/superpowers/skills/executing-plans/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/executing-plans/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/executing-plans/SKILL.md' -print -quit 2>/dev/null | grep -q .
+test -f .cursor/skills/superpowers/skills/using-git-worktrees/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/using-git-worktrees/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/using-git-worktrees/SKILL.md' -print -quit 2>/dev/null | grep -q .
+test -f .cursor/skills/superpowers/skills/verification-before-completion/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/verification-before-completion/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/verification-before-completion/SKILL.md' -print -quit 2>/dev/null | grep -q .
+test -f .cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md || test -f ~/.cursor/skills/superpowers/skills/finishing-a-development-branch/SKILL.md || find ~/.cursor/plugins/cache/cursor-public/superpowers -path '*/skills/finishing-a-development-branch/SKILL.md' -print -quit 2>/dev/null | grep -q .
 # Only if using cursor as reviewer CLI:
 # jq --version
 ```
 
 If any dependency is missing, stop and return:
 
-`Missing dependency: [specific missing item]. Install from https://github.com/obra/superpowers and copy into .cursor/skills/ or ~/.cursor/skills/, then retry.`
+`Missing dependency: [specific missing item]. Install the Cursor Superpowers plugin or install Superpowers under .cursor/skills/ or ~/.cursor/skills/, then retry.`
 
 If no plan folder exists:
 
@@ -94,6 +94,19 @@ If the user has already specified a reviewer CLI and model (e.g., "implement the
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS`. These values are fixed for the entire run.
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ### Phase 3: Set Up Worktree (REQUIRED SUB-SKILL)
 
 Invoke `superpowers:using-git-worktrees`.
@@ -272,6 +285,18 @@ Write the reviewer invocation to `/tmp/milestone-review-${REVIEW_ID}.sh` as a ba
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/milestone-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 ```bash
@@ -415,7 +440,7 @@ jq -r '.result' /tmp/milestone-review-${REVIEW_ID}.json > /tmp/milestone-review-
 ```
 
 - If `REVIEWER_CLI=codex`, extract `CODEX_SESSION_ID` from `/tmp/milestone-review-${REVIEW_ID}.runner.out` after the helper or fallback run. If the review text is only in `.runner.out`, move or copy the actual review body into `/tmp/milestone-review-${REVIEW_ID}.md` before verdict parsing.
-- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
+- If `REVIEWER_CLI=claude` or `REVIEWER_CLI=pi`, promote stdout captured by the helper or fallback runner into the markdown review file:
 
 ```bash
 cp /tmp/milestone-review-${REVIEW_ID}.runner.out /tmp/milestone-review-${REVIEW_ID}.md
@@ -467,6 +492,18 @@ If a revision contradicts the user's explicit requirements, skip it and note it
 
 Rewrite `/tmp/milestone-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call with prior-round context (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "You previously reviewed this milestone and requested revisions. Read the updated payload at /tmp/milestone-${REVIEW_ID}.md and re-review using the same ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Resume the existing session:

skills/implement-plan/opencode/SKILL.md

@@ -17,7 +17,7 @@ Required:
 - `milestone-plan.md` exists in plan folder
 - `story-tracker.md` exists in plan folder
 - Git repo with worktree support: `git worktree list`
-- OpenCode Superpowers skills symlink: `~/.config/opencode/skills/superpowers`
+- OpenCode Superpowers skills available at `~/.agents/skills/superpowers` or `~/.config/opencode/skills/superpowers`
 - Superpowers execution skills:
   - `superpowers/executing-plans`
   - `superpowers/using-git-worktrees`
@@ -27,7 +27,10 @@ Required:
 Verify before proceeding:
 
 ```bash
-ls -l ~/.config/opencode/skills/superpowers
+test -f ~/.agents/skills/superpowers/executing-plans/SKILL.md || test -f ~/.config/opencode/skills/superpowers/executing-plans/SKILL.md
+test -f ~/.agents/skills/superpowers/using-git-worktrees/SKILL.md || test -f ~/.config/opencode/skills/superpowers/using-git-worktrees/SKILL.md
+test -f ~/.agents/skills/superpowers/verification-before-completion/SKILL.md || test -f ~/.config/opencode/skills/superpowers/verification-before-completion/SKILL.md
+test -f ~/.agents/skills/superpowers/finishing-a-development-branch/SKILL.md || test -f ~/.config/opencode/skills/superpowers/finishing-a-development-branch/SKILL.md
 ```
 
 If dependencies are missing, stop immediately and return:
@@ -76,6 +79,19 @@ If the user has already specified a reviewer CLI and model (e.g., "implement the
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS`. These values are fixed for the entire run.
 
+Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`.
+
+
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
+When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the model running this workflow. If the model/provider is unavailable, surface helper stderr/status and use `pi --list-models [search]` to inspect configured models.
+
 ### Phase 4: Set Up Worktree (REQUIRED SUB-SKILL)
 
 Use OpenCode's native skill tool to load:
@@ -253,6 +269,18 @@ Write the reviewer invocation to `/tmp/milestone-review-${REVIEW_ID}.sh` as a ba
 set -euo pipefail
 ```
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call every round (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "Read the file /tmp/milestone-${REVIEW_ID}.md and review. Return exactly the required ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 ```bash
@@ -390,7 +418,7 @@ jq -r '.result' /tmp/milestone-review-${REVIEW_ID}.json > /tmp/milestone-review-
 ```
 
 - If `REVIEWER_CLI=codex`, extract `CODEX_SESSION_ID` from `/tmp/milestone-review-${REVIEW_ID}.runner.out` after the helper or fallback run. If the review text is only in `.runner.out`, move or copy the actual review body into `/tmp/milestone-review-${REVIEW_ID}.md` before verdict parsing.
-- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
+- If `REVIEWER_CLI=claude` or `REVIEWER_CLI=pi`, promote stdout captured by the helper or fallback runner into the markdown review file:
 
 ```bash
 cp /tmp/milestone-review-${REVIEW_ID}.runner.out /tmp/milestone-review-${REVIEW_ID}.md
@@ -442,6 +470,18 @@ If a revision contradicts the user's explicit requirements, skip it and note it
 
 Rewrite `/tmp/milestone-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
 
+
+**If `REVIEWER_CLI` is `pi`:**
+
+Fresh call with prior-round context (Pi reviewer calls do not use session resume):
+
+```bash
+pi --no-session --no-skills --no-prompt-templates --no-extensions --no-context-files \
+  --model "$REVIEWER_MODEL" \
+  --tools read,grep,find,ls \
+  -p "You previously reviewed this milestone and requested revisions. Read the updated payload at /tmp/milestone-${REVIEW_ID}.md and re-review using the same ## Summary, ## Findings, and ## Verdict structure."
+```
+
 **If `REVIEWER_CLI` is `codex`:**
 
 Resume the existing session:

skills/implement-plan/pi/SKILL.md

@@ -86,6 +86,14 @@ Reviewer CLI: `codex`, `claude`, `cursor`, `opencode`, `pi`, or `skip`
 
 Store `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS`.
 
+If `REVIEWER_CLI=pi`, verify the Pi reviewer binary before entering the review loop:
+
+```bash
+pi --version
+```
+
+For shorthand `pi/<pi-model-name>`, split only on the first slash when the prefix is exactly `pi`; store the complete remainder in `REVIEWER_MODEL`. Examples: `pi/claude-opus-4-7` -> `claude-opus-4-7`, `pi/anthropic/claude-opus-4-7` -> `anthropic/claude-opus-4-7`, and `pi/openrouter/anthropic/claude-opus-4-7` -> `openrouter/anthropic/claude-opus-4-7`.
+
 When `REVIEWER_CLI=pi`, the reviewer model is configured independently from the pi model running this workflow. Use any configured pi model string, including provider-qualified model IDs. If the reviewer model or provider is unavailable, surface the review helper stderr/status and ask for a configured model; use `pi --list-models [search]` to inspect configured models.
 
 The pi reviewer command rendered into `/tmp/milestone-review-${REVIEW_ID}.sh` must be isolated and read-only:

skills/web-automation/cursor/SKILL.md

@@ -0,0 +1,112 @@
+---
+name: web-automation
+description: Browse and scrape web pages using Playwright-compatible CloakBrowser. Use when automating web workflows, extracting rendered page content, handling authenticated sessions, or running multi-step browser flows.
+---
+
+# Web Automation with CloakBrowser (Cursor)
+
+Automated web browsing and scraping using Playwright-compatible CloakBrowser with two execution paths:
+
+- one-shot extraction via `extract.js`
+- broader stateful automation via `auth.ts`, `browse.ts`, `flow.ts`, `scan-local-app.ts`, and `scrape.ts`
+
+## Requirements
+
+- Node.js 20+
+- pnpm
+- Network access to download the CloakBrowser binary on first use
+
+## First-Time Setup
+
+Repo-local install:
+
+```bash
+cd .cursor/skills/web-automation/scripts
+pnpm install
+npx cloakbrowser install
+pnpm approve-builds
+pnpm rebuild better-sqlite3 esbuild
+```
+
+Global install:
+
+```bash
+cd ~/.cursor/skills/web-automation/scripts
+pnpm install
+npx cloakbrowser install
+pnpm approve-builds
+pnpm rebuild better-sqlite3 esbuild
+```
+
+## Updating CloakBrowser
+
+Run from the installed `scripts/` directory:
+
+```bash
+pnpm up cloakbrowser playwright-core
+npx cloakbrowser install
+pnpm approve-builds
+pnpm rebuild better-sqlite3 esbuild
+```
+
+## Prerequisite Check (MANDATORY)
+
+Before running automation, verify CloakBrowser and Playwright Core are installed and wired correctly.
+
+```bash
+cd .cursor/skills/web-automation/scripts || cd ~/.cursor/skills/web-automation/scripts
+node check-install.js
+```
+
+If the check fails, stop and return:
+
+"Missing dependency/config: web-automation requires `cloakbrowser` and `playwright-core` with CloakBrowser-based scripts. Run setup in this skill, then retry."
+
+If runtime fails with missing native bindings for `better-sqlite3` or `esbuild`, run:
+
+```bash
+cd .cursor/skills/web-automation/scripts || cd ~/.cursor/skills/web-automation/scripts
+pnpm approve-builds
+pnpm rebuild better-sqlite3 esbuild
+```
+
+## When To Use Which Command
+
+- Use `node extract.js "<URL>"` for a one-shot rendered fetch with JSON output.
+- Use `npx tsx scrape.ts ...` when you need markdown extraction, Readability cleanup, or selector-based scraping.
+- Use `npx tsx browse.ts ...`, `auth.ts`, or `flow.ts` when the task needs login handling, persistent sessions, clicks, typing, screenshots, or multi-step navigation.
+- Use `npx tsx scan-local-app.ts` when you need a configurable local-app smoke pass driven by `SCAN_*` and `CLOAKBROWSER_*` environment variables.
+
+## Quick Reference
+
+- Install check: `node check-install.js`
+- One-shot JSON extract: `node extract.js "https://example.com"`
+- Browse page: `npx tsx browse.ts --url "https://example.com"`
+- Scrape markdown: `npx tsx scrape.ts --url "https://example.com" --mode main --output page.md`
+- Authenticate: `npx tsx auth.ts --url "https://example.com/login"`
+- Natural-language flow: `npx tsx flow.ts --instruction 'go to https://example.com then click on "Login" then type "user@example.com" in #email then press enter'`
+- Local app smoke scan: `SCAN_BASE_URL=http://localhost:3000 SCAN_ROUTES=/,/dashboard npx tsx scan-local-app.ts`
+
+## Local App Smoke Scan
+
+`scan-local-app.ts` is intentionally generic. Configure it with environment variables instead of editing the file:
+
+- `SCAN_BASE_URL`
+- `SCAN_LOGIN_PATH`
+- `SCAN_USERNAME`
+- `SCAN_PASSWORD`
+- `SCAN_USERNAME_SELECTOR`
+- `SCAN_PASSWORD_SELECTOR`
+- `SCAN_SUBMIT_SELECTOR`
+- `SCAN_ROUTES`
+- `SCAN_REPORT_PATH`
+- `SCAN_HEADLESS`
+
+If `SCAN_USERNAME` or `SCAN_PASSWORD` are omitted, the script falls back to `CLOAKBROWSER_USERNAME` and `CLOAKBROWSER_PASSWORD`.
+
+## Notes
+
+- Sessions persist in CloakBrowser profile storage.
+- Use `--wait` for dynamic pages.
+- Use `--mode selector --selector "..."` for targeted extraction.
+- `extract.js` keeps a bounded stealth/rendered fetch path without needing a long-lived automation session.

skills/web-automation/cursor/scripts/auth.ts

@@ -0,0 +1,575 @@
+#!/usr/bin/env npx tsx
+
+/**
+ * Authentication handler for web automation
+ * Supports generic form login and Microsoft SSO (MSAL)
+ *
+ * Usage:
+ *   npx tsx auth.ts --url "https://example.com/login" --type form
+ *   npx tsx auth.ts --url "https://example.com" --type msal
+ *   npx tsx auth.ts --url "https://example.com" --type auto
+ */
+
+import { getPage, launchBrowser } from './browse.js';
+import parseArgs from 'minimist';
+import type { Page, BrowserContext } from 'playwright-core';
+import { createInterface } from 'readline';
+
+// Types
+type AuthType = 'auto' | 'form' | 'msal';
+
+interface AuthOptions {
+  url: string;
+  authType: AuthType;
+  credentials?: {
+    username: string;
+    password: string;
+  };
+  headless?: boolean;
+  timeout?: number;
+}
+
+interface AuthResult {
+  success: boolean;
+  finalUrl: string;
+  authType: AuthType;
+  message: string;
+}
+
+// Get credentials from environment or options
+function getCredentials(options?: {
+  username?: string;
+  password?: string;
+}): { username: string; password: string } | null {
+  const username = options?.username || process.env.CLOAKBROWSER_USERNAME;
+  const password = options?.password || process.env.CLOAKBROWSER_PASSWORD;
+
+  if (!username || !password) {
+    return null;
+  }
+
+  return { username, password };
+}
+
+// Prompt user for input (for MFA or credentials)
+async function promptUser(question: string, hidden = false): Promise<string> {
+  const rl = createInterface({
+    input: process.stdin,
+    output: process.stdout,
+  });
+
+  return new Promise((resolve) => {
+    if (hidden) {
+      process.stdout.write(question);
+      // Note: This is a simple implementation. For production, use a proper hidden input library
+    }
+    rl.question(question, (answer) => {
+      rl.close();
+      resolve(answer);
+    });
+  });
+}
+
+// Detect authentication type from page
+async function detectAuthType(page: Page): Promise<AuthType> {
+  const url = page.url();
+
+  // Check for Microsoft login
+  if (
+    url.includes('login.microsoftonline.com') ||
+    url.includes('login.live.com') ||
+    url.includes('login.windows.net')
+  ) {
+    return 'msal';
+  }
+
+  // Check for common form login patterns
+  const hasLoginForm = await page.evaluate(() => {
+    const passwordField = document.querySelector(
+      'input[type="password"], input[name*="password"], input[id*="password"]'
+    );
+    const usernameField = document.querySelector(
+      'input[type="email"], input[type="text"][name*="user"], input[type="text"][name*="email"], input[id*="user"], input[id*="email"]'
+    );
+    return !!(passwordField && usernameField);
+  });
+
+  if (hasLoginForm) {
+    return 'form';
+  }
+
+  return 'auto';
+}
+
+// Handle generic form login
+async function handleFormLogin(
+  page: Page,
+  credentials: { username: string; password: string },
+  timeout: number
+): Promise<boolean> {
+  console.log('Attempting form login...');
+
+  // Find and fill username/email field
+  const usernameSelectors = [
+    'input[type="email"]',
+    'input[name*="user" i]',
+    'input[name*="email" i]',
+    'input[id*="user" i]',
+    'input[id*="email" i]',
+    'input[autocomplete="username"]',
+    'input[type="text"]:first-of-type',
+  ];
+
+  let usernameField = null;
+  for (const selector of usernameSelectors) {
+    usernameField = await page.$(selector);
+    if (usernameField) break;
+  }
+
+  if (!usernameField) {
+    console.error('Could not find username/email field');
+    return false;
+  }
+
+  await usernameField.fill(credentials.username);
+  console.log('Filled username field');
+
+  // Find and fill password field
+  const passwordSelectors = [
+    'input[type="password"]',
+    'input[name*="password" i]',
+    'input[id*="password" i]',
+    'input[autocomplete="current-password"]',
+  ];
+
+  let passwordField = null;
+  for (const selector of passwordSelectors) {
+    passwordField = await page.$(selector);
+    if (passwordField) break;
+  }
+
+  if (!passwordField) {
+    console.error('Could not find password field');
+    return false;
+  }
+
+  await passwordField.fill(credentials.password);
+  console.log('Filled password field');
+
+  // Check for "Remember me" checkbox and check it
+  const rememberCheckbox = await page.$(
+    'input[type="checkbox"][name*="remember" i], input[type="checkbox"][id*="remember" i]'
+  );
+  if (rememberCheckbox) {
+    await rememberCheckbox.check();
+    console.log('Checked "Remember me" checkbox');
+  }
+
+  // Find and click submit button
+  const submitSelectors = [
+    'button[type="submit"]',
+    'input[type="submit"]',
+    'button:has-text("Sign in")',
+    'button:has-text("Log in")',
+    'button:has-text("Login")',
+    'button:has-text("Submit")',
+    '[role="button"]:has-text("Sign in")',
+  ];
+
+  let submitButton = null;
+  for (const selector of submitSelectors) {
+    submitButton = await page.$(selector);
+    if (submitButton) break;
+  }
+
+  if (!submitButton) {
+    // Try pressing Enter as fallback
+    await passwordField.press('Enter');
+  } else {
+    await submitButton.click();
+  }
+
+  console.log('Submitted login form');
+
+  // Wait for navigation or error
+  try {
+    await page.waitForNavigation({ timeout, waitUntil: 'domcontentloaded' });
+    return true;
+  } catch {
+    // Check if we're still on login page with error
+    const errorMessages = await page.$$eval(
+      '.error, .alert-danger, [role="alert"], .login-error',
+      (els) => els.map((el) => el.textContent?.trim()).filter(Boolean)
+    );
+
+    if (errorMessages.length > 0) {
+      console.error('Login error:', errorMessages.join(', '));
+      return false;
+    }
+
+    return true; // Might have succeeded without navigation
+  }
+}
+
+// Handle Microsoft SSO login
+async function handleMsalLogin(
+  page: Page,
+  credentials: { username: string; password: string },
+  timeout: number
+): Promise<boolean> {
+  console.log('Attempting Microsoft SSO login...');
+
+  const currentUrl = page.url();
+
+  // If not already on Microsoft login, wait for redirect
+  if (!currentUrl.includes('login.microsoftonline.com')) {
+    try {
+      await page.waitForURL('**/login.microsoftonline.com/**', { timeout: 10000 });
+    } catch {
+      console.log('Not redirected to Microsoft login');
+      return false;
+    }
+  }
+
+  // Wait for email input
+  const emailInput = await page.waitForSelector(
+    'input[type="email"], input[name="loginfmt"]',
+    { timeout }
+  );
+
+  if (!emailInput) {
+    console.error('Could not find email input on Microsoft login');
+    return false;
+  }
+
+  // Fill email and submit
+  await emailInput.fill(credentials.username);
+  console.log('Filled email field');
+
+  const nextButton = await page.$('input[type="submit"], button[type="submit"]');
+  if (nextButton) {
+    await nextButton.click();
+  } else {
+    await emailInput.press('Enter');
+  }
+
+  // Wait for password page
+  try {
+    await page.waitForSelector(
+      'input[type="password"], input[name="passwd"]',
+      { timeout }
+    );
+  } catch {
+    // Might be using passwordless auth or different flow
+    console.log('Password field not found - might be using different auth flow');
+    return false;
+  }
+
+  // Fill password
+  const passwordInput = await page.$('input[type="password"], input[name="passwd"]');
+  if (!passwordInput) {
+    console.error('Could not find password input');
+    return false;
+  }
+
+  await passwordInput.fill(credentials.password);
+  console.log('Filled password field');
+
+  // Submit
+  const signInButton = await page.$('input[type="submit"], button[type="submit"]');
+  if (signInButton) {
+    await signInButton.click();
+  } else {
+    await passwordInput.press('Enter');
+  }
+
+  // Handle "Stay signed in?" prompt
+  try {
+    const staySignedInButton = await page.waitForSelector(
+      'input[value="Yes"], button:has-text("Yes")',
+      { timeout: 5000 }
+    );
+    if (staySignedInButton) {
+      await staySignedInButton.click();
+      console.log('Clicked "Stay signed in" button');
+    }
+  } catch {
+    // Prompt might not appear
+  }
+
+  // Check for Conditional Access Policy error
+  const caError = await page.$('text=Conditional Access policy');
+  if (caError) {
+    console.error('Blocked by Conditional Access Policy');
+    // Take screenshot for debugging
+    await page.screenshot({ path: 'ca-policy-error.png' });
+    console.log('Screenshot saved: ca-policy-error.png');
+    return false;
+  }
+
+  // Wait for redirect away from Microsoft login
+  try {
+    await page.waitForURL(
+      (url) => !url.href.includes('login.microsoftonline.com'),
+      { timeout }
+    );
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+// Check if user is already authenticated
+async function isAuthenticated(page: Page, targetUrl: string): Promise<boolean> {
+  const currentUrl = page.url();
+
+  // If we're on the target URL (not a login page), we're likely authenticated
+  if (currentUrl.startsWith(targetUrl)) {
+    // Check for common login page indicators
+    const isLoginPage = await page.evaluate(() => {
+      const loginIndicators = [
+        'input[type="password"]',
+        'form[action*="login"]',
+        'form[action*="signin"]',
+        '.login-form',
+        '#login',
+      ];
+      return loginIndicators.some((sel) => document.querySelector(sel) !== null);
+    });
+
+    return !isLoginPage;
+  }
+
+  return false;
+}
+
+// Main authentication function
+export async function authenticate(options: AuthOptions): Promise<AuthResult> {
+  const browser = await launchBrowser({ headless: options.headless ?? true });
+  const page = await browser.newPage();
+  const timeout = options.timeout ?? 30000;
+
+  try {
+    // Navigate to URL
+    console.log(`Navigating to: ${options.url}`);
+    await page.goto(options.url, { timeout: 60000, waitUntil: 'domcontentloaded' });
+
+    // Check if already authenticated
+    if (await isAuthenticated(page, options.url)) {
+      return {
+        success: true,
+        finalUrl: page.url(),
+        authType: 'auto',
+        message: 'Already authenticated (session persisted from profile)',
+      };
+    }
+
+    // Get credentials
+    const credentials = options.credentials
+      ? options.credentials
+      : getCredentials();
+
+    if (!credentials) {
+      // No credentials - open interactive browser
+      console.log('\nNo credentials provided. Opening browser for manual login...');
+      console.log('Please complete the login process manually.');
+      console.log('The session will be saved to your profile.');
+
+      // Switch to headed mode for manual login
+      await browser.close();
+      const interactiveBrowser = await launchBrowser({ headless: false });
+      const interactivePage = await interactiveBrowser.newPage();
+      await interactivePage.goto(options.url);
+
+      await promptUser('\nPress Enter when you have completed login...');
+
+      const finalUrl = interactivePage.url();
+      await interactiveBrowser.close();
+
+      return {
+        success: true,
+        finalUrl,
+        authType: 'auto',
+        message: 'Manual login completed - session saved to profile',
+      };
+    }
+
+    // Detect auth type if auto
+    let authType = options.authType;
+    if (authType === 'auto') {
+      authType = await detectAuthType(page);
+      console.log(`Detected auth type: ${authType}`);
+    }
+
+    // Handle authentication based on type
+    let success = false;
+    switch (authType) {
+      case 'msal':
+        success = await handleMsalLogin(page, credentials, timeout);
+        break;
+      case 'form':
+      default:
+        success = await handleFormLogin(page, credentials, timeout);
+        break;
+    }
+
+    const finalUrl = page.url();
+
+    return {
+      success,
+      finalUrl,
+      authType,
+      message: success
+        ? `Authentication successful - session saved to profile`
+        : 'Authentication failed',
+    };
+  } finally {
+    await browser.close();
+  }
+}
+
+// Navigate to authenticated page (handles auth if needed)
+export async function navigateAuthenticated(
+  url: string,
+  options?: {
+    credentials?: { username: string; password: string };
+    headless?: boolean;
+  }
+): Promise<{ page: Page; browser: BrowserContext }> {
+  const { page, browser } = await getPage({ headless: options?.headless ?? true });
+
+  await page.goto(url, { timeout: 60000, waitUntil: 'domcontentloaded' });
+
+  // Check if we need to authenticate
+  if (!(await isAuthenticated(page, url))) {
+    console.log('Session expired or not authenticated. Attempting login...');
+
+    // Get credentials
+    const credentials = options?.credentials ?? getCredentials();
+
+    if (!credentials) {
+      throw new Error(
+        'Authentication required but no credentials provided. ' +
+          'Set CLOAKBROWSER_USERNAME and CLOAKBROWSER_PASSWORD environment variables.'
+      );
+    }
+
+    // Detect and handle auth
+    const authType = await detectAuthType(page);
+
+    let success = false;
+    if (authType === 'msal') {
+      success = await handleMsalLogin(page, credentials, 30000);
+    } else {
+      success = await handleFormLogin(page, credentials, 30000);
+    }
+
+    if (!success) {
+      await browser.close();
+      throw new Error('Authentication failed');
+    }
+
+    // Navigate back to original URL if we were redirected
+    if (!page.url().startsWith(url)) {
+      await page.goto(url, { timeout: 60000, waitUntil: 'domcontentloaded' });
+    }
+  }
+
+  return { page, browser };
+}
+
+// CLI entry point
+async function main() {
+  const args = parseArgs(process.argv.slice(2), {
+    string: ['url', 'type', 'username', 'password'],
+    boolean: ['headless', 'help'],
+    default: {
+      type: 'auto',
+      headless: false, // Default to headed for auth so user can see/interact
+    },
+    alias: {
+      u: 'url',
+      t: 'type',
+      h: 'help',
+    },
+  });
+
+  if (args.help || !args.url) {
+    console.log(`
+Web Authentication Handler
+
+Usage:
+  npx tsx auth.ts --url <url> [options]
+
+Options:
+  -u, --url <url>         URL to authenticate (required)
+  -t, --type <type>       Auth type: auto, form, or msal (default: auto)
+  --username <user>       Username/email (or set CLOAKBROWSER_USERNAME env var)
+  --password <pass>       Password (or set CLOAKBROWSER_PASSWORD env var)
+  --headless <bool>       Run in headless mode (default: false for auth)
+  -h, --help              Show this help message
+
+Auth Types:
+  auto    Auto-detect authentication type
+  form    Generic username/password form
+  msal    Microsoft SSO (login.microsoftonline.com)
+
+Environment Variables:
+  CLOAKBROWSER_USERNAME       Default username/email for authentication
+  CLOAKBROWSER_PASSWORD       Default password for authentication
+
+Examples:
+  # Interactive login (no credentials, opens browser)
+  npx tsx auth.ts --url "https://example.com/login"
+
+  # Form login with credentials
+  npx tsx auth.ts --url "https://example.com/login" --type form \\
+    --username "user@example.com" --password "secret"
+
+  # Microsoft SSO login
+  CLOAKBROWSER_USERNAME=user@company.com CLOAKBROWSER_PASSWORD=secret \\
+  npx tsx auth.ts --url "https://internal.company.com" --type msal
+
+Notes:
+  - Session is saved to ~/.cloakbrowser-profile/ for persistence
+  - After successful auth, subsequent browses will be authenticated
+  - Use --headless false if you need to handle MFA manually
+`);
+    process.exit(args.help ? 0 : 1);
+  }
+
+  const authType = args.type as AuthType;
+  if (!['auto', 'form', 'msal'].includes(authType)) {
+    console.error(`Invalid auth type: ${authType}. Must be auto, form, or msal.`);
+    process.exit(1);
+  }
+
+  try {
+    const result = await authenticate({
+      url: args.url,
+      authType,
+      credentials:
+        args.username && args.password
+          ? { username: args.username, password: args.password }
+          : undefined,
+      headless: args.headless,
+    });
+
+    console.log(`\nAuthentication result:`);
+    console.log(`  Success: ${result.success}`);
+    console.log(`  Auth type: ${result.authType}`);
+    console.log(`  Final URL: ${result.finalUrl}`);
+    console.log(`  Message: ${result.message}`);
+
+    process.exit(result.success ? 0 : 1);
+  } catch (error) {
+    console.error('Error:', error instanceof Error ? error.message : error);
+    process.exit(1);
+  }
+}
+
+// Run if executed directly
+const isMainModule = process.argv[1]?.includes('auth.ts');
+if (isMainModule) {
+  main();
+}

skills/web-automation/cursor/scripts/browse.ts

@@ -0,0 +1,188 @@
+#!/usr/bin/env npx tsx
+
+/**
+ * Browser launcher using CloakBrowser with persistent profile
+ *
+ * Usage:
+ *   npx tsx browse.ts --url "https://example.com"
+ *   npx tsx browse.ts --url "https://example.com" --screenshot --output page.png
+ *   npx tsx browse.ts --url "https://example.com" --headless false --wait 5000
+ */
+
+import { launchPersistentContext } from 'cloakbrowser';
+import { homedir } from 'os';
+import { join } from 'path';
+import { existsSync, mkdirSync } from 'fs';
+import parseArgs from 'minimist';
+import type { Page, BrowserContext } from 'playwright-core';
+
+interface BrowseOptions {
+  url: string;
+  headless?: boolean;
+  screenshot?: boolean;
+  output?: string;
+  wait?: number;
+  timeout?: number;
+  interactive?: boolean;
+}
+
+interface BrowseResult {
+  title: string;
+  url: string;
+  screenshotPath?: string;
+}
+
+function sleep(ms: number): Promise<void> {
+  return new Promise((resolve) => setTimeout(resolve, ms));
+}
+
+const getProfilePath = (): string => {
+  const customPath = process.env.CLOAKBROWSER_PROFILE_PATH;
+  if (customPath) return customPath;
+
+  const profileDir = join(homedir(), '.cloakbrowser-profile');
+  if (!existsSync(profileDir)) {
+    mkdirSync(profileDir, { recursive: true });
+  }
+  return profileDir;
+};
+
+export async function launchBrowser(options: {
+  headless?: boolean;
+}): Promise<BrowserContext> {
+  const profilePath = getProfilePath();
+  const envHeadless = process.env.CLOAKBROWSER_HEADLESS;
+  const headless = options.headless ?? (envHeadless ? envHeadless === 'true' : true);
+
+  console.log(`Using profile: ${profilePath}`);
+  console.log(`Headless mode: ${headless}`);
+
+  const context = await launchPersistentContext({
+    userDataDir: profilePath,
+    headless,
+    humanize: true,
+  });
+
+  return context;
+}
+
+export async function browse(options: BrowseOptions): Promise<BrowseResult> {
+  const browser = await launchBrowser({ headless: options.headless });
+  const page = browser.pages()[0] || await browser.newPage();
+
+  try {
+    console.log(`Navigating to: ${options.url}`);
+    await page.goto(options.url, {
+      timeout: options.timeout ?? 60000,
+      waitUntil: 'domcontentloaded',
+    });
+
+    if (options.wait) {
+      console.log(`Waiting ${options.wait}ms...`);
+      await sleep(options.wait);
+    }
+
+    const result: BrowseResult = {
+      title: await page.title(),
+      url: page.url(),
+    };
+
+    console.log(`Page title: ${result.title}`);
+    console.log(`Final URL: ${result.url}`);
+
+    if (options.screenshot) {
+      const outputPath = options.output ?? 'screenshot.png';
+      await page.screenshot({ path: outputPath, fullPage: true });
+      result.screenshotPath = outputPath;
+      console.log(`Screenshot saved: ${outputPath}`);
+    }
+
+    if (options.interactive) {
+      console.log('\nInteractive mode - browser will stay open.');
+      console.log('Press Ctrl+C to close.');
+      await new Promise(() => {});
+    }
+
+    return result;
+  } finally {
+    if (!options.interactive) {
+      await browser.close();
+    }
+  }
+}
+
+export async function getPage(options?: {
+  headless?: boolean;
+}): Promise<{ page: Page; browser: BrowserContext }> {
+  const browser = await launchBrowser({ headless: options?.headless });
+  const page = browser.pages()[0] || await browser.newPage();
+  return { page, browser };
+}
+
+async function main() {
+  const args = parseArgs(process.argv.slice(2), {
+    string: ['url', 'output'],
+    boolean: ['screenshot', 'headless', 'interactive', 'help'],
+    default: {
+      headless: true,
+      screenshot: false,
+      interactive: false,
+    },
+    alias: {
+      u: 'url',
+      o: 'output',
+      s: 'screenshot',
+      h: 'help',
+      i: 'interactive',
+    },
+  });
+
+  if (args.help || !args.url) {
+    console.log(`
+Web Browser with CloakBrowser
+
+Usage:
+  npx tsx browse.ts --url <url> [options]
+
+Options:
+  -u, --url <url>       URL to navigate to (required)
+  -s, --screenshot      Take a screenshot of the page
+  -o, --output <path>   Output path for screenshot (default: screenshot.png)
+  --headless <bool>     Run in headless mode (default: true)
+  --wait <ms>           Wait time after page load in milliseconds
+  --timeout <ms>        Navigation timeout (default: 60000)
+  -i, --interactive     Keep browser open for manual interaction
+  -h, --help            Show this help message
+
+Examples:
+  npx tsx browse.ts --url "https://example.com"
+  npx tsx browse.ts --url "https://example.com" --screenshot --output page.png
+  npx tsx browse.ts --url "https://example.com" --headless false --interactive
+
+Environment Variables:
+  CLOAKBROWSER_PROFILE_PATH  Custom profile directory (default: ~/.cloakbrowser-profile/)
+  CLOAKBROWSER_HEADLESS      Default headless mode (true/false)
+`);
+    process.exit(args.help ? 0 : 1);
+  }
+
+  try {
+    await browse({
+      url: args.url,
+      headless: args.headless,
+      screenshot: args.screenshot,
+      output: args.output,
+      wait: args.wait ? parseInt(args.wait, 10) : undefined,
+      timeout: args.timeout ? parseInt(args.timeout, 10) : undefined,
+      interactive: args.interactive,
+    });
+  } catch (error) {
+    console.error('Error:', error instanceof Error ? error.message : error);
+    process.exit(1);
+  }
+}
+
+const isMainModule = process.argv[1]?.includes('browse.ts');
+if (isMainModule) {
+  main();
+}

skills/web-automation/cursor/scripts/check-install.js

@@ -0,0 +1,40 @@
+#!/usr/bin/env node
+
+import fs from "node:fs";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = path.dirname(__filename);
+
+function fail(message, details) {
+  const payload = { error: message };
+  if (details) payload.details = details;
+  process.stderr.write(`${JSON.stringify(payload)}\n`);
+  process.exit(1);
+}
+
+async function main() {
+  try {
+    await import("cloakbrowser");
+    await import("playwright-core");
+  } catch (error) {
+    fail(
+      "Missing dependency/config: web-automation requires cloakbrowser and playwright-core.",
+      error instanceof Error ? error.message : String(error)
+    );
+  }
+
+  const browsePath = path.join(__dirname, "browse.ts");
+  const browseSource = fs.readFileSync(browsePath, "utf8");
+  if (!/launchPersistentContext/.test(browseSource) || !/from ['"]cloakbrowser['"]/.test(browseSource)) {
+    fail("browse.ts is not configured for CloakBrowser.");
+  }
+
+  process.stdout.write("OK: cloakbrowser + playwright-core installed\n");
+  process.stdout.write("OK: CloakBrowser integration detected in browse.ts\n");
+}
+
+main().catch((error) => {
+  fail("Install check failed.", error instanceof Error ? error.message : String(error));
+});

skills/web-automation/cursor/scripts/extract.js

@@ -0,0 +1,188 @@
+#!/usr/bin/env node
+
+import fs from "node:fs";
+import path from "node:path";
+import { fileURLToPath } from "node:url";
+
+const DEFAULT_WAIT_MS = 5000;
+const MAX_WAIT_MS = 20000;
+const NAV_TIMEOUT_MS = 30000;
+const EXTRA_CHALLENGE_WAIT_MS = 8000;
+const CONTENT_LIMIT = 12000;
+
+const __filename = fileURLToPath(import.meta.url);
+const __dirname = path.dirname(__filename);
+
+function fail(message, details) {
+  const payload = { error: message };
+  if (details) payload.details = details;
+  process.stderr.write(`${JSON.stringify(payload)}\n`);
+  process.exit(1);
+}
+
+function parseWaitTime(raw) {
+  const value = Number.parseInt(raw || `${DEFAULT_WAIT_MS}`, 10);
+  if (!Number.isFinite(value) || value < 0) return DEFAULT_WAIT_MS;
+  return Math.min(value, MAX_WAIT_MS);
+}
+
+function parseTarget(rawUrl) {
+  if (!rawUrl) {
+    fail("Missing URL. Usage: node extract.js <URL>");
+  }
+
+  let parsed;
+  try {
+    parsed = new URL(rawUrl);
+  } catch (error) {
+    fail("Invalid URL.", error.message);
+  }
+
+  if (!["http:", "https:"].includes(parsed.protocol)) {
+    fail("Only http and https URLs are allowed.");
+  }
+
+  return parsed.toString();
+}
+
+function ensureParentDir(filePath) {
+  if (!filePath) return;
+  fs.mkdirSync(path.dirname(filePath), { recursive: true });
+}
+
+function sleep(ms) {
+  return new Promise((resolve) => setTimeout(resolve, ms));
+}
+
+async function detectChallenge(page) {
+  try {
+    return await page.evaluate(() => {
+      const text = (document.body?.innerText || "").toLowerCase();
+      return (
+        text.includes("checking your browser") ||
+        text.includes("just a moment") ||
+        text.includes("verify you are human") ||
+        text.includes("press and hold") ||
+        document.querySelector('iframe[src*="challenge"]') !== null ||
+        document.querySelector('iframe[src*="cloudflare"]') !== null
+      );
+    });
+  } catch {
+    return false;
+  }
+}
+
+async function loadCloakBrowser() {
+  try {
+    return await import("cloakbrowser");
+  } catch (error) {
+    fail(
+      "CloakBrowser is not installed for this skill. Run pnpm install in this skill's scripts directory first.",
+      error.message
+    );
+  }
+}
+
+async function runWithStderrLogs(fn) {
+  const originalLog = console.log;
+  const originalError = console.error;
+  console.log = (...args) => process.stderr.write(`${args.join(" ")}\n`);
+  console.error = (...args) => process.stderr.write(`${args.join(" ")}\n`);
+  try {
+    return await fn();
+  } finally {
+    console.log = originalLog;
+    console.error = originalError;
+  }
+}
+
+async function main() {
+  const requestedUrl = parseTarget(process.argv[2]);
+  const waitTime = parseWaitTime(process.env.WAIT_TIME);
+  const screenshotPath = process.env.SCREENSHOT_PATH || "";
+  const saveHtml = process.env.SAVE_HTML === "true";
+  const headless = process.env.HEADLESS !== "false";
+  const userAgent = process.env.USER_AGENT || undefined;
+  const startedAt = Date.now();
+  const { ensureBinary, launchContext } = await loadCloakBrowser();
+
+  let context;
+  try {
+    await runWithStderrLogs(() => ensureBinary());
+
+    context = await runWithStderrLogs(() => launchContext({
+      headless,
+      userAgent,
+      locale: "en-US",
+      viewport: { width: 1440, height: 900 },
+      humanize: true,
+    }));
+
+    const page = await context.newPage();
+    const response = await page.goto(requestedUrl, {
+      waitUntil: "domcontentloaded",
+      timeout: NAV_TIMEOUT_MS
+    });
+
+    await sleep(waitTime);
+
+    let challengeDetected = await detectChallenge(page);
+    if (challengeDetected) {
+      await sleep(EXTRA_CHALLENGE_WAIT_MS);
+      challengeDetected = await detectChallenge(page);
+    }
+
+    const extracted = await page.evaluate((contentLimit) => {
+      const bodyText = document.body?.innerText || "";
+      return {
+        finalUrl: window.location.href,
+        title: document.title || "",
+        content: bodyText.slice(0, contentLimit),
+        metaDescription:
+          document.querySelector('meta[name="description"]')?.content ||
+          document.querySelector('meta[property="og:description"]')?.content ||
+          ""
+      };
+    }, CONTENT_LIMIT);
+
+    const result = {
+      requestedUrl,
+      finalUrl: extracted.finalUrl,
+      title: extracted.title,
+      content: extracted.content,
+      metaDescription: extracted.metaDescription,
+      status: response ? response.status() : null,
+      challengeDetected,
+      elapsedSeconds: ((Date.now() - startedAt) / 1000).toFixed(2)
+    };
+
+    if (screenshotPath) {
+      ensureParentDir(screenshotPath);
+      await page.screenshot({ path: screenshotPath, fullPage: false, timeout: 10000 });
+      result.screenshot = screenshotPath;
+    }
+
+    if (saveHtml) {
+      const htmlTarget = screenshotPath
+        ? screenshotPath.replace(/\.[^.]+$/, ".html")
+        : path.resolve(__dirname, `page-${Date.now()}.html`);
+      ensureParentDir(htmlTarget);
+      fs.writeFileSync(htmlTarget, await page.content());
+      result.htmlFile = htmlTarget;
+    }
+
+    process.stdout.write(`${JSON.stringify(result, null, 2)}\n`);
+    await context.close();
+  } catch (error) {
+    if (context) {
+      try {
+        await context.close();
+      } catch {
+        // Ignore close errors after the primary failure.
+      }
+    }
+    fail("Scrape failed.", error.message);
+  }
+}
+
+main();

skills/web-automation/cursor/scripts/flow.ts

@@ -0,0 +1,329 @@
+#!/usr/bin/env npx tsx
+
+import parseArgs from 'minimist';
+import type { Page } from 'playwright-core';
+import { launchBrowser } from './browse';
+
+type Step =
+  | { action: 'goto'; url: string }
+  | { action: 'click'; selector?: string; text?: string; role?: string; name?: string }
+  | { action: 'type'; selector?: string; text: string }
+  | { action: 'press'; key: string; selector?: string }
+  | { action: 'wait'; ms: number }
+  | { action: 'screenshot'; path: string }
+  | { action: 'extract'; selector: string; count?: number };
+
+function normalizeNavigationUrl(rawUrl: string): string {
+  let parsed: URL;
+  try {
+    parsed = new URL(rawUrl);
+  } catch {
+    throw new Error(`Invalid navigation URL: ${rawUrl}`);
+  }
+
+  if (!['http:', 'https:'].includes(parsed.protocol)) {
+    throw new Error(`Only http and https URLs are allowed in flow steps: ${rawUrl}`);
+  }
+
+  return parsed.toString();
+}
+
+function normalizeKey(k: string): string {
+  if (!k) return 'Enter';
+  const lower = k.toLowerCase();
+  if (lower === 'enter' || lower === 'return') return 'Enter';
+  if (lower === 'tab') return 'Tab';
+  if (lower === 'escape' || lower === 'esc') return 'Escape';
+  return k;
+}
+
+function splitInstructions(instruction: string): string[] {
+  return instruction
+    .split(/\bthen\b|;/gi)
+    .map((s) => s.trim())
+    .filter(Boolean);
+}
+
+function parseInstruction(instruction: string): Step[] {
+  const parts = splitInstructions(instruction);
+  const steps: Step[] = [];
+
+  for (const p of parts) {
+    // go to https://...
+    const goto = p.match(/^(?:go to|open|navigate to)\s+(https?:\/\/\S+)/i);
+    if (goto) {
+      steps.push({ action: 'goto', url: normalizeNavigationUrl(goto[1]) });
+      continue;
+    }
+
+    // click on "text" or click #selector or click button "name"
+    const clickRole = p.match(/^click\s+(button|link|textbox|img|image|tab)\s+"([^"]+)"$/i);
+    if (clickRole) {
+      const role = clickRole[1].toLowerCase() === 'image' ? 'img' : clickRole[1].toLowerCase();
+      steps.push({ action: 'click', role, name: clickRole[2] });
+      continue;
+    }
+    const clickText = p.match(/^click(?: on)?\s+"([^"]+)"/i);
+    if (clickText) {
+      steps.push({ action: 'click', text: clickText[1] });
+      continue;
+    }
+    const clickSelector = p.match(/^click(?: on)?\s+(#[\w-]+|\.[\w-]+|[a-z]+\[[^\]]+\])/i);
+    if (clickSelector) {
+      steps.push({ action: 'click', selector: clickSelector[1] });
+      continue;
+    }
+
+    // type "text" [in selector]
+    const typeInto = p.match(/^type\s+"([^"]+)"\s+in\s+(.+)$/i);
+    if (typeInto) {
+      steps.push({ action: 'type', text: typeInto[1], selector: typeInto[2].trim() });
+      continue;
+    }
+    const typeOnly = p.match(/^type\s+"([^"]+)"$/i);
+    if (typeOnly) {
+      steps.push({ action: 'type', text: typeOnly[1] });
+      continue;
+    }
+
+    // press enter [in selector]
+    const pressIn = p.match(/^press\s+(\w+)\s+in\s+(.+)$/i);
+    if (pressIn) {
+      steps.push({ action: 'press', key: normalizeKey(pressIn[1]), selector: pressIn[2].trim() });
+      continue;
+    }
+    const pressOnly = p.match(/^press\s+(\w+)$/i);
+    if (pressOnly) {
+      steps.push({ action: 'press', key: normalizeKey(pressOnly[1]) });
+      continue;
+    }
+
+    // wait 2s / wait 500ms
+    const waitS = p.match(/^wait\s+(\d+)\s*s(?:ec(?:onds?)?)?$/i);
+    if (waitS) {
+      steps.push({ action: 'wait', ms: parseInt(waitS[1], 10) * 1000 });
+      continue;
+    }
+    const waitMs = p.match(/^wait\s+(\d+)\s*ms$/i);
+    if (waitMs) {
+      steps.push({ action: 'wait', ms: parseInt(waitMs[1], 10) });
+      continue;
+    }
+
+    // screenshot path
+    const shot = p.match(/^screenshot(?: to)?\s+(.+)$/i);
+    if (shot) {
+      steps.push({ action: 'screenshot', path: shot[1].trim() });
+      continue;
+    }
+
+    throw new Error(`Could not parse step: "${p}"`);
+  }
+
+  return steps;
+}
+
+function validateSteps(steps: Step[]): Step[] {
+  return steps.map((step) =>
+    step.action === 'goto'
+      ? {
+          ...step,
+          url: normalizeNavigationUrl(step.url),
+        }
+      : step
+  );
+}
+
+function escapeRegExp(value: string): string {
+  return value.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+}
+
+function isLikelyLoginText(text: string): boolean {
+  return /(login|accedi|sign\s*in|entra)/i.test(text);
+}
+
+async function clickByText(page: Page, text: string): Promise<boolean> {
+  const patterns = [new RegExp(`^${escapeRegExp(text)}$`, 'i'), new RegExp(escapeRegExp(text), 'i')];
+
+  for (const pattern of patterns) {
+    const targets = [
+      page.getByRole('button', { name: pattern }).first(),
+      page.getByRole('link', { name: pattern }).first(),
+      page.getByText(pattern).first(),
+    ];
+
+    for (const target of targets) {
+      if (await target.count()) {
+        try {
+          await target.click({ timeout: 8000 });
+          return true;
+        } catch {
+          // keep trying next candidate
+        }
+      }
+    }
+  }
+
+  return false;
+}
+
+async function fallbackLoginNavigation(page: Page, requestedText: string): Promise<boolean> {
+  if (!isLikelyLoginText(requestedText)) return false;
+
+  const current = new URL(page.url());
+
+  const candidateLinks = await page.evaluate(() => {
+    const loginTerms = ['login', 'accedi', 'sign in', 'entra'];
+    const anchors = Array.from(document.querySelectorAll('a[href], a[onclick], button[onclick]')) as Array<HTMLAnchorElement | HTMLButtonElement>;
+
+    return anchors
+      .map((el) => {
+        const text = (el.textContent || '').trim().toLowerCase();
+        const href = (el as HTMLAnchorElement).getAttribute('href') || '';
+        return { text, href };
+      })
+      .filter((x) => x.text && loginTerms.some((t) => x.text.includes(t)))
+      .map((x) => x.href)
+      .filter(Boolean);
+  });
+
+  // Prefer real URLs (not javascript:)
+  const realCandidate = candidateLinks.find((h) => /login|account\/login/i.test(h) && !h.startsWith('javascript:'));
+  if (realCandidate) {
+    const target = new URL(realCandidate, page.url()).toString();
+    await page.goto(target, { waitUntil: 'domcontentloaded', timeout: 60000 });
+    return true;
+  }
+
+  // Site-specific fallback for Corriere
+  if (/corriere\.it$/i.test(current.hostname) || /\.corriere\.it$/i.test(current.hostname)) {
+    await page.goto('https://www.corriere.it/account/login', {
+      waitUntil: 'domcontentloaded',
+      timeout: 60000,
+    });
+    return true;
+  }
+
+  return false;
+}
+
+async function typeInBestTarget(page: Page, text: string, selector?: string) {
+  if (selector) {
+    await page.locator(selector).first().click({ timeout: 10000 });
+    await page.locator(selector).first().fill(text);
+    return;
+  }
+  const loc = page.locator('input[name="q"], input[type="search"], input[type="text"], textarea').first();
+  await loc.click({ timeout: 10000 });
+  await loc.fill(text);
+}
+
+async function pressOnTarget(page: Page, key: string, selector?: string) {
+  if (selector) {
+    await page.locator(selector).first().press(key);
+    return;
+  }
+  await page.keyboard.press(key);
+}
+
+async function runSteps(page: Page, steps: Step[]) {
+  for (const step of steps) {
+    switch (step.action) {
+      case 'goto':
+        await page.goto(normalizeNavigationUrl(step.url), {
+          waitUntil: 'domcontentloaded',
+          timeout: 60000,
+        });
+        break;
+      case 'click':
+        if (step.selector) {
+          await page.locator(step.selector).first().click({ timeout: 15000 });
+        } else if (step.role && step.name) {
+          await page.getByRole(step.role as any, { name: new RegExp(escapeRegExp(step.name), 'i') }).first().click({ timeout: 15000 });
+        } else if (step.text) {
+          const clicked = await clickByText(page, step.text);
+          if (!clicked) {
+            const recovered = await fallbackLoginNavigation(page, step.text);
+            if (!recovered) {
+              throw new Error(`Could not click target text: ${step.text}`);
+            }
+          }
+        } else {
+          throw new Error('click step missing selector/text/role');
+        }
+        try {
+          await page.waitForLoadState('domcontentloaded', { timeout: 10000 });
+        } catch {
+          // no navigation is fine
+        }
+        break;
+      case 'type':
+        await typeInBestTarget(page, step.text, step.selector);
+        break;
+      case 'press':
+        await pressOnTarget(page, step.key, step.selector);
+        break;
+      case 'wait':
+        await page.waitForTimeout(step.ms);
+        break;
+      case 'screenshot':
+        await page.screenshot({ path: step.path, fullPage: true });
+        break;
+      case 'extract': {
+        const items = await page.locator(step.selector).allTextContents();
+        const out = items.slice(0, step.count ?? items.length).map((t) => t.trim()).filter(Boolean);
+        console.log(JSON.stringify(out, null, 2));
+        break;
+      }
+      default:
+        throw new Error('Unknown step');
+    }
+  }
+}
+
+async function main() {
+  const args = parseArgs(process.argv.slice(2), {
+    string: ['instruction', 'steps'],
+    boolean: ['headless', 'help'],
+    default: { headless: true },
+    alias: { i: 'instruction', s: 'steps', h: 'help' },
+  });
+
+  if (args.help || (!args.instruction && !args.steps)) {
+    console.log(`
+General Web Flow Runner (CloakBrowser)
+
+Usage:
+  npx tsx flow.ts --instruction "go to https://example.com then type \"hello\" then press enter"
+  npx tsx flow.ts --steps '[{"action":"goto","url":"https://example.com"}]'
+
+Supported natural steps:
+  - go to/open/navigate to <url>
+  - click on "Text"
+  - click <css-selector>
+  - type "text"
+  - type "text" in <css-selector>
+  - press <key>
+  - press <key> in <css-selector>
+  - wait <N>s | wait <N>ms
+  - screenshot <path>
+`);
+    process.exit(args.help ? 0 : 1);
+  }
+
+  const steps = validateSteps(args.steps ? JSON.parse(args.steps) : parseInstruction(args.instruction));
+  const browser = await launchBrowser({ headless: args.headless });
+  const page = await browser.newPage();
+
+  try {
+    await runSteps(page, steps);
+    console.log('Flow complete. Final URL:', page.url());
+  } finally {
+    await browser.close();
+  }
+}
+
+main().catch((e) => {
+  console.error('Error:', e instanceof Error ? e.message : e);
+  process.exit(1);
+});

skills/web-automation/cursor/scripts/package.json

@@ -0,0 +1,36 @@
+{
+  "name": "web-automation-scripts",
+  "version": "1.0.0",
+  "description": "Web browsing and scraping scripts using CloakBrowser",
+  "type": "module",
+  "scripts": {
+    "check-install": "node check-install.js",
+    "extract": "node extract.js",
+    "browse": "tsx browse.ts",
+    "auth": "tsx auth.ts",
+    "flow": "tsx flow.ts",
+    "scrape": "tsx scrape.ts",
+    "typecheck": "tsc --noEmit -p tsconfig.json",
+    "lint": "pnpm run typecheck && node --check check-install.js && node --check extract.js",
+    "fetch-browser": "npx cloakbrowser install"
+  },
+  "dependencies": {
+    "@mozilla/readability": "^0.5.0",
+    "better-sqlite3": "^12.6.2",
+    "cloakbrowser": "^0.3.22",
+    "jsdom": "^24.0.0",
+    "minimist": "^1.2.8",
+    "playwright-core": "^1.59.1",
+    "turndown": "^7.1.2",
+    "turndown-plugin-gfm": "^1.0.2"
+  },
+  "devDependencies": {
+    "@types/jsdom": "^21.1.6",
+    "@types/minimist": "^1.2.5",
+    "@types/turndown": "^5.0.4",
+    "esbuild": "0.27.0",
+    "tsx": "^4.7.0",
+    "typescript": "^5.3.0"
+  },
+  "packageManager": "pnpm@10.18.1+sha512.77a884a165cbba2d8d1c19e3b4880eee6d2fcabd0d879121e282196b80042351d5eb3ca0935fa599da1dc51265cc68816ad2bddd2a2de5ea9fdf92adbec7cd34"
+}

skills/web-automation/cursor/scripts/scan-local-app.ts

@@ -0,0 +1,174 @@
+#!/usr/bin/env npx tsx
+
+import { mkdirSync, writeFileSync } from 'fs';
+import { dirname, resolve } from 'path';
+import { getPage } from './browse.js';
+
+type NavResult = {
+  requestedUrl: string;
+  url: string;
+  status: number | null;
+  title: string;
+  error?: string;
+};
+
+type RouteCheck = {
+  route: string;
+  result: NavResult;
+  heading: string | null;
+};
+
+const DEFAULT_BASE_URL = 'http://localhost:3000';
+const DEFAULT_REPORT_PATH = resolve(process.cwd(), 'scan-local-app.md');
+
+function env(name: string): string | undefined {
+  const value = process.env[name]?.trim();
+  return value ? value : undefined;
+}
+
+function getRoutes(baseUrl: string): string[] {
+  const routeList = env('SCAN_ROUTES');
+  if (routeList) {
+    return routeList
+      .split(',')
+      .map((route) => route.trim())
+      .filter(Boolean)
+      .map((route) => new URL(route, baseUrl).toString());
+  }
+
+  return [baseUrl];
+}
+
+async function gotoWithStatus(page: any, url: string): Promise<NavResult> {
+  const response = await page
+    .goto(url, { waitUntil: 'domcontentloaded', timeout: 60000 })
+    .catch((error: unknown) => ({ error }));
+
+  if (response?.error) {
+    return {
+      requestedUrl: url,
+      url: page.url(),
+      status: null,
+      title: await page.title().catch(() => ''),
+      error: String(response.error),
+    };
+  }
+
+  return {
+    requestedUrl: url,
+    url: page.url(),
+    status: response ? response.status() : null,
+    title: await page.title().catch(() => ''),
+  };
+}
+
+async function textOrNull(page: any, selector: string): Promise<string | null> {
+  const locator = page.locator(selector).first();
+  try {
+    if ((await locator.count()) === 0) return null;
+    const value = await locator.textContent();
+    return value ? value.trim().replace(/\s+/g, ' ') : null;
+  } catch {
+    return null;
+  }
+}
+
+async function loginIfConfigured(page: any, baseUrl: string, lines: string[]) {
+  const loginPath = env('SCAN_LOGIN_PATH');
+  const username = env('SCAN_USERNAME') ?? env('CLOAKBROWSER_USERNAME');
+  const password = env('SCAN_PASSWORD') ?? env('CLOAKBROWSER_PASSWORD');
+  const usernameSelector = env('SCAN_USERNAME_SELECTOR') ?? 'input[type="email"], input[name="email"]';
+  const passwordSelector = env('SCAN_PASSWORD_SELECTOR') ?? 'input[type="password"], input[name="password"]';
+  const submitSelector = env('SCAN_SUBMIT_SELECTOR') ?? 'button[type="submit"], input[type="submit"]';
+
+  if (!loginPath) {
+    lines.push('## Login');
+    lines.push('- Skipped: set `SCAN_LOGIN_PATH` to enable login smoke checks.');
+    lines.push('');
+    return;
+  }
+
+  const loginUrl = new URL(loginPath, baseUrl).toString();
+  lines.push('## Login');
+  lines.push(`- Login URL: ${loginUrl}`);
+  await gotoWithStatus(page, loginUrl);
+
+  if (!username || !password) {
+    lines.push('- Skipped: set `SCAN_USERNAME`/`SCAN_PASSWORD` or `CLOAKBROWSER_USERNAME`/`CLOAKBROWSER_PASSWORD`.');
+    lines.push('');
+    return;
+  }
+
+  await page.locator(usernameSelector).first().fill(username);
+  await page.locator(passwordSelector).first().fill(password);
+  await page.locator(submitSelector).first().click();
+  await page.waitForTimeout(2500);
+
+  lines.push(`- After submit URL: ${page.url()}`);
+  lines.push(`- Cookie count: ${(await page.context().cookies()).length}`);
+  lines.push('');
+}
+
+async function checkRoutes(page: any, baseUrl: string, lines: string[]) {
+  const routes = getRoutes(baseUrl);
+  const routeChecks: RouteCheck[] = [];
+
+  for (const url of routes) {
+    const result = await gotoWithStatus(page, url);
+    const heading = await textOrNull(page, 'h1');
+    routeChecks.push({
+      route: url,
+      result,
+      heading,
+    });
+  }
+
+  lines.push('## Route Checks');
+  for (const check of routeChecks) {
+    const relativeUrl = check.route.startsWith(baseUrl) ? check.route.slice(baseUrl.length) || '/' : check.route;
+    const finalPath = check.result.url.startsWith(baseUrl)
+      ? check.result.url.slice(baseUrl.length) || '/'
+      : check.result.url;
+    const suffix = check.heading ? `, h1="${check.heading}"` : '';
+    const errorSuffix = check.result.error ? `, error="${check.result.error}"` : '';
+    lines.push(
+      `- ${relativeUrl} → status ${check.result.status ?? 'ERR'} (final ${finalPath})${suffix}${errorSuffix}`
+    );
+  }
+  lines.push('');
+}
+
+async function main() {
+  const baseUrl = env('SCAN_BASE_URL') ?? DEFAULT_BASE_URL;
+  const reportPath = resolve(env('SCAN_REPORT_PATH') ?? DEFAULT_REPORT_PATH);
+  const headless = (env('SCAN_HEADLESS') ?? env('CLOAKBROWSER_HEADLESS') ?? 'true') === 'true';
+  const { page, browser } = await getPage({ headless });
+  const lines: string[] = [];
+
+  lines.push('# Web Automation Scan (local)');
+  lines.push('');
+  lines.push(`- Base URL: ${baseUrl}`);
+  lines.push(`- Timestamp: ${new Date().toISOString()}`);
+  lines.push(`- Headless: ${headless}`);
+  lines.push(`- Report Path: ${reportPath}`);
+  lines.push('');
+
+  try {
+    await loginIfConfigured(page, baseUrl, lines);
+    await checkRoutes(page, baseUrl, lines);
+    lines.push('## Notes');
+    lines.push('- This generic smoke helper records route availability and top-level headings for a local app.');
+    lines.push('- Configure login and route coverage with `SCAN_*` environment variables.');
+  } finally {
+    await browser.close();
+  }
+
+  mkdirSync(dirname(reportPath), { recursive: true });
+  writeFileSync(reportPath, `${lines.join('\n')}\n`, 'utf-8');
+  console.log(`Report written to ${reportPath}`);
+}
+
+main().catch((error) => {
+  console.error(error);
+  process.exitCode = 1;
+});

skills/web-automation/cursor/scripts/scrape.ts

@@ -0,0 +1,351 @@
+#!/usr/bin/env npx tsx
+
+/**
+ * Web scraper that extracts content to markdown
+ *
+ * Usage:
+ *   npx tsx scrape.ts --url "https://example.com" --mode main
+ *   npx tsx scrape.ts --url "https://example.com" --mode full --output page.md
+ *   npx tsx scrape.ts --url "https://example.com" --mode selector --selector ".content"
+ */
+
+import TurndownService from 'turndown';
+import * as turndownPluginGfm from 'turndown-plugin-gfm';
+import { Readability } from '@mozilla/readability';
+import { JSDOM } from 'jsdom';
+import { writeFileSync } from 'fs';
+import parseArgs from 'minimist';
+import { getPage } from './browse.js';
+
+// Types
+type ScrapeMode = 'main' | 'full' | 'selector';
+
+interface ScrapeOptions {
+  url: string;
+  mode: ScrapeMode;
+  selector?: string;
+  output?: string;
+  includeLinks?: boolean;
+  includeTables?: boolean;
+  includeImages?: boolean;
+  headless?: boolean;
+  wait?: number;
+}
+
+interface ScrapeResult {
+  title: string;
+  url: string;
+  markdown: string;
+  byline?: string;
+  excerpt?: string;
+}
+
+// Configure Turndown for markdown conversion
+function createTurndownService(options: {
+  includeLinks?: boolean;
+  includeTables?: boolean;
+  includeImages?: boolean;
+}): TurndownService {
+  const turndown = new TurndownService({
+    headingStyle: 'atx',
+    hr: '---',
+    bulletListMarker: '-',
+    codeBlockStyle: 'fenced',
+    fence: '```',
+    emDelimiter: '*',
+    strongDelimiter: '**',
+    linkStyle: 'inlined',
+  });
+
+  // Add GFM support (tables, strikethrough, task lists)
+  turndown.use(turndownPluginGfm.gfm);
+
+  // Custom rule for code blocks with language detection
+  turndown.addRule('codeBlockWithLanguage', {
+    filter: (node) => {
+      return (
+        node.nodeName === 'PRE' &&
+        node.firstChild?.nodeName === 'CODE'
+      );
+    },
+    replacement: (_content, node) => {
+      const codeNode = node.firstChild as HTMLElement;
+      const className = codeNode.getAttribute('class') || '';
+      const langMatch = className.match(/language-(\w+)/);
+      const lang = langMatch ? langMatch[1] : '';
+      const code = codeNode.textContent || '';
+      return `\n\n\`\`\`${lang}\n${code}\n\`\`\`\n\n`;
+    },
+  });
+
+  // Remove images if not included
+  if (!options.includeImages) {
+    turndown.addRule('removeImages', {
+      filter: 'img',
+      replacement: () => '',
+    });
+  }
+
+  // Remove links but keep text if not included
+  if (!options.includeLinks) {
+    turndown.addRule('removeLinks', {
+      filter: 'a',
+      replacement: (content) => content,
+    });
+  }
+
+  // Remove script, style, nav, footer, aside elements
+  turndown.remove(['script', 'style', 'nav', 'footer', 'aside', 'noscript']);
+
+  return turndown;
+}
+
+// Extract main content using Readability
+function extractMainContent(html: string, url: string): {
+  content: string;
+  title: string;
+  byline?: string;
+  excerpt?: string;
+} {
+  const dom = new JSDOM(html, { url });
+  const reader = new Readability(dom.window.document);
+  const article = reader.parse();
+
+  if (!article) {
+    throw new Error('Could not extract main content from page');
+  }
+
+  return {
+    content: article.content,
+    title: article.title,
+    byline: article.byline || undefined,
+    excerpt: article.excerpt || undefined,
+  };
+}
+
+// Scrape a URL and return markdown
+export async function scrape(options: ScrapeOptions): Promise<ScrapeResult> {
+  const { page, browser } = await getPage({ headless: options.headless ?? true });
+
+  try {
+    // Navigate to URL
+    console.log(`Navigating to: ${options.url}`);
+    await page.goto(options.url, {
+      timeout: 60000,
+      waitUntil: 'domcontentloaded',
+    });
+
+    // Wait if specified
+    if (options.wait) {
+      console.log(`Waiting ${options.wait}ms for dynamic content...`);
+      await page.waitForTimeout(options.wait);
+    }
+
+    const pageTitle = await page.title();
+    const pageUrl = page.url();
+
+    let html: string;
+    let title = pageTitle;
+    let byline: string | undefined;
+    let excerpt: string | undefined;
+
+    // Get HTML based on mode
+    switch (options.mode) {
+      case 'main': {
+        // Get full page HTML and extract with Readability
+        const fullHtml = await page.content();
+        const extracted = extractMainContent(fullHtml, pageUrl);
+        html = extracted.content;
+        title = extracted.title || pageTitle;
+        byline = extracted.byline;
+        excerpt = extracted.excerpt;
+        break;
+      }
+
+      case 'selector': {
+        if (!options.selector) {
+          throw new Error('Selector mode requires --selector option');
+        }
+        const element = await page.$(options.selector);
+        if (!element) {
+          throw new Error(`Selector not found: ${options.selector}`);
+        }
+        html = await element.innerHTML();
+        break;
+      }
+
+      case 'full':
+      default: {
+        // Get body content, excluding common non-content elements
+        html = await page.evaluate(() => {
+          // Remove common non-content elements
+          const selectorsToRemove = [
+            'script', 'style', 'noscript', 'iframe',
+            'nav', 'header', 'footer', '.cookie-banner',
+            '.advertisement', '.ads', '#ads', '.social-share',
+            '.comments', '#comments', '.sidebar'
+          ];
+
+          selectorsToRemove.forEach(selector => {
+            document.querySelectorAll(selector).forEach(el => el.remove());
+          });
+
+          return document.body.innerHTML;
+        });
+        break;
+      }
+    }
+
+    // Convert to markdown
+    const turndown = createTurndownService({
+      includeLinks: options.includeLinks ?? true,
+      includeTables: options.includeTables ?? true,
+      includeImages: options.includeImages ?? false,
+    });
+
+    let markdown = turndown.turndown(html);
+
+    // Add title as H1 if not already present
+    if (!markdown.startsWith('# ')) {
+      markdown = `# ${title}\n\n${markdown}`;
+    }
+
+    // Add metadata header
+    const metadataLines = [
+      `<!-- Scraped from: ${pageUrl} -->`,
+      byline ? `<!-- Author: ${byline} -->` : null,
+      excerpt ? `<!-- Excerpt: ${excerpt} -->` : null,
+      `<!-- Scraped at: ${new Date().toISOString()} -->`,
+      '',
+    ].filter(Boolean);
+
+    markdown = metadataLines.join('\n') + '\n' + markdown;
+
+    // Clean up excessive whitespace
+    markdown = markdown
+      .replace(/\n{4,}/g, '\n\n\n')
+      .replace(/[ \t]+$/gm, '')
+      .trim();
+
+    const result: ScrapeResult = {
+      title,
+      url: pageUrl,
+      markdown,
+      byline,
+      excerpt,
+    };
+
+    // Save to file if output specified
+    if (options.output) {
+      writeFileSync(options.output, markdown, 'utf-8');
+      console.log(`Markdown saved to: ${options.output}`);
+    }
+
+    return result;
+  } finally {
+    await browser.close();
+  }
+}
+
+// CLI entry point
+async function main() {
+  const args = parseArgs(process.argv.slice(2), {
+    string: ['url', 'mode', 'selector', 'output'],
+    boolean: ['headless', 'links', 'tables', 'images', 'help'],
+    default: {
+      mode: 'main',
+      headless: true,
+      links: true,
+      tables: true,
+      images: false,
+    },
+    alias: {
+      u: 'url',
+      m: 'mode',
+      s: 'selector',
+      o: 'output',
+      h: 'help',
+    },
+  });
+
+  if (args.help || !args.url) {
+    console.log(`
+Web Scraper - Extract content to Markdown
+
+Usage:
+  npx tsx scrape.ts --url <url> [options]
+
+Options:
+  -u, --url <url>       URL to scrape (required)
+  -m, --mode <mode>     Scrape mode: main, full, or selector (default: main)
+  -s, --selector <sel>  CSS selector for selector mode
+  -o, --output <path>   Output file path for markdown
+  --headless <bool>     Run in headless mode (default: true)
+  --wait <ms>           Wait time for dynamic content
+  --links               Include links in output (default: true)
+  --tables              Include tables in output (default: true)
+  --images              Include images in output (default: false)
+  -h, --help            Show this help message
+
+Scrape Modes:
+  main      Extract main article content using Readability (best for articles)
+  full      Full page content with common elements removed
+  selector  Extract specific element by CSS selector
+
+Examples:
+  npx tsx scrape.ts --url "https://docs.example.com/guide" --mode main
+  npx tsx scrape.ts --url "https://example.com" --mode full --output page.md
+  npx tsx scrape.ts --url "https://example.com" --mode selector --selector ".api-docs"
+  npx tsx scrape.ts --url "https://example.com" --mode main --no-links --output clean.md
+
+Output Format:
+  - GitHub Flavored Markdown (tables, strikethrough, task lists)
+  - Proper heading hierarchy
+  - Code blocks with language detection
+  - Metadata comments at top (source URL, date)
+`);
+    process.exit(args.help ? 0 : 1);
+  }
+
+  const mode = args.mode as ScrapeMode;
+  if (!['main', 'full', 'selector'].includes(mode)) {
+    console.error(`Invalid mode: ${mode}. Must be main, full, or selector.`);
+    process.exit(1);
+  }
+
+  try {
+    const result = await scrape({
+      url: args.url,
+      mode,
+      selector: args.selector,
+      output: args.output,
+      includeLinks: args.links,
+      includeTables: args.tables,
+      includeImages: args.images,
+      headless: args.headless,
+      wait: args.wait ? parseInt(args.wait, 10) : undefined,
+    });
+
+    // Print result summary
+    console.log(`\nScrape complete:`);
+    console.log(`  Title: ${result.title}`);
+    console.log(`  URL: ${result.url}`);
+    if (result.byline) console.log(`  Author: ${result.byline}`);
+    console.log(`  Markdown length: ${result.markdown.length} chars`);
+
+    // Print markdown if not saved to file
+    if (!args.output) {
+      console.log('\n--- Markdown Output ---\n');
+      console.log(result.markdown);
+    }
+  } catch (error) {
+    console.error('Error:', error instanceof Error ? error.message : error);
+    process.exit(1);
+  }
+}
+
+// Run if executed directly
+const isMainModule = process.argv[1]?.includes('scrape.ts');
+if (isMainModule) {
+  main();
+}

skills/web-automation/cursor/scripts/test-full.ts

@@ -0,0 +1,36 @@
+import { launchPersistentContext } from 'cloakbrowser';
+import { homedir } from 'os';
+import { join } from 'path';
+import { mkdirSync, existsSync } from 'fs';
+
+async function test() {
+  const profilePath = join(homedir(), '.cloakbrowser-profile');
+  if (!existsSync(profilePath)) {
+    mkdirSync(profilePath, { recursive: true });
+  }
+
+  console.log('Profile path:', profilePath);
+  console.log('Launching CloakBrowser with full options...');
+
+  const browser = await launchPersistentContext({
+    headless: true,
+    userDataDir: profilePath,
+    humanize: true,
+  });
+
+  console.log('Browser launched');
+  const page = browser.pages()[0] || await browser.newPage();
+  console.log('Page created');
+
+  await page.goto('https://github.com', { timeout: 30000 });
+  console.log('Navigated to:', page.url());
+  console.log('Title:', await page.title());
+
+  await page.screenshot({ path: '/tmp/github-test.png' });
+  console.log('Screenshot saved');
+
+  await browser.close();
+  console.log('Done');
+}
+
+test().catch(console.error);

skills/web-automation/cursor/scripts/test-minimal.ts

@@ -0,0 +1,23 @@
+import { launch } from 'cloakbrowser';
+
+async function test() {
+  console.log('Launching CloakBrowser with minimal config...');
+
+  const browser = await launch({
+    headless: true,
+    humanize: true,
+  });
+
+  console.log('Browser launched');
+  const page = await browser.newPage();
+  console.log('Page created');
+
+  await page.goto('https://example.com', { timeout: 30000 });
+  console.log('Navigated to:', page.url());
+  console.log('Title:', await page.title());
+
+  await browser.close();
+  console.log('Done');
+}
+
+test().catch(console.error);

skills/web-automation/cursor/scripts/test-profile.ts

@@ -0,0 +1,33 @@
+import { launchPersistentContext } from 'cloakbrowser';
+import { homedir } from 'os';
+import { join } from 'path';
+import { mkdirSync, existsSync } from 'fs';
+
+async function test() {
+  const profilePath = join(homedir(), '.cloakbrowser-profile');
+  if (!existsSync(profilePath)) {
+    mkdirSync(profilePath, { recursive: true });
+  }
+
+  console.log('Profile path:', profilePath);
+  console.log('Launching with persistent userDataDir...');
+
+  const browser = await launchPersistentContext({
+    headless: true,
+    userDataDir: profilePath,
+    humanize: true,
+  });
+
+  console.log('Browser launched');
+  const page = browser.pages()[0] || await browser.newPage();
+  console.log('Page created');
+
+  await page.goto('https://example.com', { timeout: 30000 });
+  console.log('Navigated to:', page.url());
+  console.log('Title:', await page.title());
+
+  await browser.close();
+  console.log('Done');
+}
+
+test().catch(console.error);

skills/web-automation/cursor/scripts/tsconfig.json

@@ -0,0 +1,16 @@
+{
+  "compilerOptions": {
+    "target": "ES2022",
+    "module": "ESNext",
+    "moduleResolution": "bundler",
+    "esModuleInterop": true,
+    "allowSyntheticDefaultImports": true,
+    "strict": true,
+    "skipLibCheck": true,
+    "resolveJsonModule": true,
+    "outDir": "./dist",
+    "rootDir": "."
+  },
+  "include": ["*.ts"],
+  "exclude": ["node_modules", "dist"]
+}

skills/web-automation/cursor/scripts/turndown-plugin-gfm.d.ts

@@ -0,0 +1,8 @@
+declare module 'turndown-plugin-gfm' {
+  import TurndownService from 'turndown';
+
+  export function gfm(turndownService: TurndownService): void;
+  export function strikethrough(turndownService: TurndownService): void;
+  export function tables(turndownService: TurndownService): void;
+  export function taskListItems(turndownService: TurndownService): void;
+}

skills/web-automation/scripts/sync-variants.sh

@@ -5,6 +5,7 @@ ROOT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)
 SOURCE_DIR="${ROOT_DIR}/codex/scripts"
 TARGETS=(
   "claude-code"
+  "cursor"
   "opencode"
   "pi"
 )