Align reviewer runtime and Telegram notifications

This commit is contained in:
Stefano Fiorini
2026-03-24 11:45:58 -05:00
parent 4d37674626
commit 63a048a26c
17 changed files with 1756 additions and 200 deletions

View File

@@ -18,6 +18,7 @@ Create structured implementation plans with milestone and story tracking, and op
- 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`
- Telegram notification setup is documented in [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md)
If dependencies are missing, stop and return:
@@ -115,10 +116,13 @@ Verify Superpowers dependencies exist in your agent skills root:
- Creates plans under `ai_plan/YYYY-MM-DD-<short-title>/`.
- Ensures `/ai_plan/` is in `.gitignore`.
- Commits `.gitignore` update locally when added.
- Asks which reviewer CLI and model to use (or accepts `skip` for no review).
- Iteratively reviews the plan with the chosen reviewer (max 5 rounds) before generating files.
- Asks which reviewer CLI, model, and max rounds to use (or accepts `skip` for no review).
- Iteratively reviews the plan with the chosen reviewer (default max 10 rounds) before generating files.
- Runs reviewer commands through `reviewer-runtime/run-review.sh` when available, with fallback to direct synchronous execution only if the helper is missing.
- Waits as long as the reviewer runtime keeps emitting per-minute `In progress N` heartbeats.
- Requires reviewer findings to be ordered `P0` through `P3`, with `P3` explicitly non-blocking.
- Captures reviewer stderr and helper status logs for diagnostics and retains them on failed, empty-output, or operator-decision review rounds.
- Sends completion notifications through Telegram only when the shared setup in [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md) is installed and configured.
- Produces:
- `original-plan.md`
- `final-transcript.md`
@@ -130,13 +134,24 @@ 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`) and model, or skips
1. **Configure** — user picks a reviewer CLI (`codex`, `claude`, `cursor`), 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
5. **Revise** — the planning agent addresses each issue and re-submits
6. **Repeat**up to 5 rounds until the reviewer returns `VERDICT: APPROVED`
7. **Finalize** — approved plan is used to generate the plan file package
4. **Feedback** — reviewer evaluates correctness, risks, missing steps, alternatives, security, and returns `## Summary`, `## Findings`, and `## Verdict`
5. **Prioritize** — findings are ordered `P0`, `P1`, `P2`, `P3`
6. **Revise**the planning agent addresses findings in priority order and re-submits
7. **Repeat** — up to max rounds until the reviewer returns `VERDICT: APPROVED`
8. **Finalize** — approved plan is used to generate the plan file package
### Reviewer Output Contract
- `P0` = total blocker
- `P1` = major risk
- `P2` = must-fix before approval
- `P3` = cosmetic / nice to have
- Each severity section should use `- None.` when empty
- `VERDICT: APPROVED` is valid only when no `P0`, `P1`, or `P2` findings remain
- `P3` findings are non-blocking, but the caller should still try to fix them when cheap and safe
### Runtime Artifacts
@@ -153,16 +168,18 @@ The review flow may create these temp artifacts:
Status log lines use this format:
```text
ts=<ISO-8601> level=<info|warn|error> state=<running-silent|running-active|stall-warning|completed|completed-empty-output|failed|needs-operator-decision> elapsed_s=<int> pid=<int> stdout_bytes=<int> stderr_bytes=<int> note="<short message>"
ts=<ISO-8601> level=<info|warn|error> state=<running-silent|running-active|in-progress|stall-warning|completed|completed-empty-output|failed|needs-operator-decision> elapsed_s=<int> pid=<int> stdout_bytes=<int> stderr_bytes=<int> note="<short message>"
```
`stall-warning` is a heartbeat/status-log state only. It is not a terminal review result.
`in-progress` is the liveness heartbeat emitted roughly once per minute with `note="In progress N"`.
`stall-warning` is a non-terminal status-log state only. It does not mean the caller should stop waiting if `in-progress` heartbeats continue.
### Failure Handling
- `completed-empty-output` means the reviewer exited without producing review text; surface `.stderr` and `.status`, then retry only after diagnosing the cause.
- `needs-operator-decision` means the helper reached hard-timeout escalation; surface `.status` and decide whether to keep waiting, abort, or retry with different parameters.
- `needs-operator-decision` means the helper reached hard-timeout escalation; surface `.status` and decide whether to extend the timeout, abort, or retry with different parameters.
- Successful rounds clean up temp artifacts. Failed, empty-output, and operator-decision rounds should retain `.stderr`, `.status`, and `.runner.out` until diagnosed.
- As long as fresh `in-progress` heartbeats continue to arrive roughly once per minute, the caller should keep waiting.
### Supported Reviewer CLIs
@@ -178,6 +195,12 @@ For all three 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
## Notifications
- Telegram is the only supported completion notification path.
- Shared setup: [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md)
- Notification failures are non-blocking, but they must be surfaced to the user.
## Template Guardrails
All plan templates now include guardrail sections that enforce:

View File

@@ -25,6 +25,7 @@ Execute an existing plan (created by `create-plan`) in an isolated git worktree,
- 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`
- Telegram notification setup is documented in [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md)
If dependencies are missing, stop and return:
@@ -133,10 +134,13 @@ Verify Superpowers execution dependencies exist in your agent skills root:
- Runs lint/typecheck/tests as a gate before each milestone review.
- Sends each milestone to a reviewer CLI for approval (max rounds configurable, default 10).
- Runs reviewer commands through `reviewer-runtime/run-review.sh` when available, with fallback to direct synchronous execution only if the helper is missing.
- Waits as long as the reviewer runtime keeps emitting per-minute `In progress N` heartbeats.
- Requires reviewer findings to be ordered `P0` through `P3`, with `P3` explicitly non-blocking.
- Captures reviewer stderr and helper status logs for diagnostics and retains them on failed, empty-output, or operator-decision review rounds.
- Commits each milestone locally only after reviewer approval (does not push).
- After all milestones approved, merges worktree branch to parent and deletes worktree.
- Supports resume: detects existing worktree and `in-dev`/`completed` stories.
- Sends completion notifications through Telegram only when the shared setup in [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md) is installed and configured.
## Milestone Review Loop
@@ -145,11 +149,22 @@ After each milestone is implemented and verified, the skill sends it to a second
1. **Configure** — user picks a reviewer CLI (`codex`, `claude`, `cursor`) 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
5. **Revise** — the implementing agent addresses each issue, re-verifies, and re-submits
6. **Repeat**up to max rounds (default 10) until the reviewer returns `VERDICT: APPROVED`
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`
### Reviewer Output Contract
- `P0` = total blocker
- `P1` = major risk
- `P2` = must-fix before approval
- `P3` = cosmetic / nice to have
- Each severity section should use `- None.` when empty
- `VERDICT: APPROVED` is valid only when no `P0`, `P1`, or `P2` findings remain
- `P3` findings are non-blocking, but the caller should still try to fix them when cheap and safe
### Runtime Artifacts
The milestone review flow may create these temp artifacts:
@@ -165,16 +180,18 @@ The milestone review flow may create these temp artifacts:
Status log lines use this format:
```text
ts=<ISO-8601> level=<info|warn|error> state=<running-silent|running-active|stall-warning|completed|completed-empty-output|failed|needs-operator-decision> elapsed_s=<int> pid=<int> stdout_bytes=<int> stderr_bytes=<int> note="<short message>"
ts=<ISO-8601> level=<info|warn|error> state=<running-silent|running-active|in-progress|stall-warning|completed|completed-empty-output|failed|needs-operator-decision> elapsed_s=<int> pid=<int> stdout_bytes=<int> stderr_bytes=<int> note="<short message>"
```
`stall-warning` is a heartbeat/status-log state only. It is not a terminal review result.
`in-progress` is the liveness heartbeat emitted roughly once per minute with `note="In progress N"`.
`stall-warning` is a non-terminal status-log state only. It does not mean the caller should stop waiting if `in-progress` heartbeats continue.
### Failure Handling
- `completed-empty-output` means the reviewer exited without producing review text; surface `.stderr` and `.status`, then retry only after diagnosing the cause.
- `needs-operator-decision` means the helper reached hard-timeout escalation; surface `.status` and decide whether to keep waiting, abort, or retry with different parameters.
- `needs-operator-decision` means the helper reached hard-timeout escalation; surface `.status` and decide whether to extend the timeout, abort, or retry with different parameters.
- Successful rounds clean up temp artifacts. Failed, empty-output, and operator-decision rounds should retain `.stderr`, `.status`, and `.runner.out` until diagnosed.
- As long as fresh `in-progress` heartbeats continue to arrive roughly once per minute, the caller should keep waiting.
### Supported Reviewer CLIs
@@ -190,6 +207,12 @@ For all three 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
## Notifications
- Telegram is the only supported completion notification path.
- Shared setup: [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md)
- Notification failures are non-blocking, but they must be surfaced to the user.
The helper also supports manual override flags for diagnostics:
```bash

View File

@@ -7,6 +7,7 @@ This directory contains user-facing docs for each skill.
- [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.
- [IMPLEMENT-PLAN.md](./IMPLEMENT-PLAN.md) — Includes requirements, install, verification, and milestone review workflow for implement-plan.
- [TELEGRAM-NOTIFICATIONS.md](./TELEGRAM-NOTIFICATIONS.md) — Shared Telegram notification setup used by reviewer-driven skills.
- [WEB-AUTOMATION.md](./WEB-AUTOMATION.md) — Includes requirements, install, dependency verification, and usage examples for web-automation.
## Repo Setup

View File

@@ -0,0 +1,96 @@
# TELEGRAM-NOTIFICATIONS
## Purpose
Shared setup for Telegram completion notifications used by reviewer-driven skills such as `create-plan` and `implement-plan`.
## Requirements
- Telegram bot token in `TELEGRAM_BOT_TOKEN`
- Telegram chat id in `TELEGRAM_CHAT_ID`
- Notification helper installed beside the shared reviewer runtime:
- Codex: `~/.codex/skills/reviewer-runtime/notify-telegram.sh`
- Claude Code: `~/.claude/skills/reviewer-runtime/notify-telegram.sh`
- OpenCode: `~/.config/opencode/skills/reviewer-runtime/notify-telegram.sh`
- Cursor: `.cursor/skills/reviewer-runtime/notify-telegram.sh` or `~/.cursor/skills/reviewer-runtime/notify-telegram.sh`
## Install
The helper ships from `skills/reviewer-runtime/` together with `run-review.sh`.
### Codex
```bash
mkdir -p ~/.codex/skills/reviewer-runtime
cp -R skills/reviewer-runtime/* ~/.codex/skills/reviewer-runtime/
```
### Claude Code
```bash
mkdir -p ~/.claude/skills/reviewer-runtime
cp -R skills/reviewer-runtime/* ~/.claude/skills/reviewer-runtime/
```
### OpenCode
```bash
mkdir -p ~/.config/opencode/skills/reviewer-runtime
cp -R skills/reviewer-runtime/* ~/.config/opencode/skills/reviewer-runtime/
```
### Cursor
Repo-local install:
```bash
mkdir -p .cursor/skills/reviewer-runtime
cp -R skills/reviewer-runtime/* .cursor/skills/reviewer-runtime/
```
Global install:
```bash
mkdir -p ~/.cursor/skills/reviewer-runtime
cp -R skills/reviewer-runtime/* ~/.cursor/skills/reviewer-runtime/
```
## Verify Installation
```bash
test -x ~/.codex/skills/reviewer-runtime/notify-telegram.sh || true
test -x ~/.claude/skills/reviewer-runtime/notify-telegram.sh || true
test -x ~/.config/opencode/skills/reviewer-runtime/notify-telegram.sh || true
test -x .cursor/skills/reviewer-runtime/notify-telegram.sh || test -x ~/.cursor/skills/reviewer-runtime/notify-telegram.sh || true
```
## Configure Telegram
Export the required variables before running a skill that sends completion notifications:
```bash
export TELEGRAM_BOT_TOKEN="<bot-token>"
export TELEGRAM_CHAT_ID="<chat-id>"
```
Optional:
```bash
export TELEGRAM_API_BASE_URL="https://api.telegram.org"
```
## Test the Helper
Example:
```bash
TELEGRAM_BOT_TOKEN="<bot-token>" \
TELEGRAM_CHAT_ID="<chat-id>" \
skills/reviewer-runtime/notify-telegram.sh --message "Telegram notification test"
```
## Rules
- Telegram is the only supported completion notification path for these skills.
- Notification failures are non-blocking, but they must be surfaced to the user.
- Skills should report when Telegram is not configured instead of silently pretending a notification was sent.

View File

@@ -47,7 +47,10 @@ If the user has already specified a reviewer CLI and model (e.g., "create a plan
- For `cursor`: **run `cursor-agent models` first** to see your account's available models (availability varies by subscription)
- Accept any model string the user provides
Store the chosen `REVIEWER_CLI` and `REVIEWER_MODEL` for Phase 6 (Iterative Plan Review).
3. **Max review rounds for the plan?** (default: 10)
- If the user does not provide a value, set `MAX_ROUNDS=10`.
Store the chosen `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phase 6 (Iterative Plan Review).
### Phase 4: Design (REQUIRED SUB-SKILL)
- Invoke `superpowers:brainstorming` explicitly.
@@ -61,7 +64,7 @@ Store the chosen `REVIEWER_CLI` and `REVIEWER_MODEL` for Phase 6 (Iterative Plan
### Phase 6: Iterative Plan Review
Send the plan to the configured reviewer CLI for feedback. Revise and re-submit until approved (max 5 rounds).
Send the plan to the configured reviewer CLI for feedback. Revise and re-submit until approved (default max 10 rounds).
**Skip this phase entirely if reviewer was set to `skip`.**
@@ -86,10 +89,60 @@ Resolve the shared reviewer helper from the installed Claude Code skills directo
REVIEWER_RUNTIME=~/.claude/skills/reviewer-runtime/run-review.sh
```
Set helper success-artifact args before writing the command script:
```bash
HELPER_SUCCESS_FILE_ARGS=()
case "$REVIEWER_CLI" in
codex)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/plan-review-${REVIEW_ID}.md)
;;
cursor)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/plan-review-${REVIEW_ID}.json)
;;
esac
```
#### Step 2: Write Plan to Temp File
Write the complete plan (milestones, stories, design decisions, specs) to `/tmp/plan-${REVIEW_ID}.md`.
#### Review Contract (Applies to Every Round)
The reviewer response must use this structure:
```text
## Summary
...
## Findings
### P0
- ...
### P1
- ...
### P2
- ...
### P3
- ...
## Verdict
VERDICT: APPROVED
```
Rules:
- Order findings from `P0` to `P3`.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- Use `- None.` when a severity has no findings.
- `VERDICT: APPROVED` is allowed only when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking.
- The calling agent should still try to fix `P3` findings when they are cheap and safe.
#### Liveness Contract (Applies While Review Is Running)
- The shared reviewer runtime emits `state=in-progress note="In progress N"` heartbeats every 60 seconds while the reviewer child is alive.
- The calling agent must keep waiting as long as a fresh `In progress N` heartbeat keeps arriving roughly once per minute.
- Do not abort just because the review is slow, a soft timeout fired, or a `stall-warning` line appears, as long as the `In progress N` heartbeat continues.
- Treat missing heartbeats, `state=failed`, `state=completed-empty-output`, and `state=needs-operator-decision` as escalation signals.
#### Step 3: Submit to Reviewer (Round 1)
Write the reviewer invocation to `/tmp/plan-review-${REVIEW_ID}.sh` as a bash script:
@@ -113,8 +166,21 @@ codex exec \
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE"
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking."
```
Do not try to capture the Codex session ID yet. When using the helper, extract it from `/tmp/plan-review-${REVIEW_ID}.runner.out` after the command completes (look for `session id: <uuid>`), then store it as `CODEX_SESSION_ID` for resume in subsequent rounds.
@@ -133,8 +199,21 @@ $(cat /tmp/plan-${REVIEW_ID}.md)
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -155,8 +234,21 @@ cursor-agent -p \
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
> /tmp/plan-review-${REVIEW_ID}.json
```
@@ -170,13 +262,16 @@ if [ -x "$REVIEWER_RUNTIME" ]; then
--command-file /tmp/plan-review-${REVIEW_ID}.sh \
--stdout-file /tmp/plan-review-${REVIEW_ID}.runner.out \
--stderr-file /tmp/plan-review-${REVIEW_ID}.stderr \
--status-file /tmp/plan-review-${REVIEW_ID}.status
--status-file /tmp/plan-review-${REVIEW_ID}.status \
"${HELPER_SUCCESS_FILE_ARGS[@]}"
else
echo "Warning: reviewer runtime helper not found at $REVIEWER_RUNTIME; falling back to direct synchronous review." >&2
bash /tmp/plan-review-${REVIEW_ID}.sh >/tmp/plan-review-${REVIEW_ID}.runner.out 2>/tmp/plan-review-${REVIEW_ID}.stderr
fi
```
Run the helper in the foreground and watch its live stdout for `state=in-progress` heartbeats. If your agent environment buffers command output until exit, start the helper in the background and poll `/tmp/plan-review-${REVIEW_ID}.status` separately instead of treating heartbeats as post-hoc-only data.
After the command completes:
- If `REVIEWER_CLI=cursor`, extract the final review text:
@@ -186,6 +281,13 @@ jq -r '.result' /tmp/plan-review-${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-${REVIEW_ID}.md` before verdict parsing.
- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
```bash
cp /tmp/plan-review-${REVIEW_ID}.runner.out /tmp/plan-review-${REVIEW_ID}.md
```
Fallback is allowed only when the helper is missing or not executable.
#### Step 4: Read Review & Check Verdict
@@ -202,17 +304,19 @@ jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_I
[Reviewer feedback]
```
3. Check verdict:
- **VERDICT: APPROVED** → proceed to Phase 7 (Initialize workspace)
- **VERDICT: REVISE** → go to Step 5
- No clear verdict but positive / no actionable items → treat as approved
4. While the reviewer is still running, keep waiting as long as fresh `state=in-progress note="In progress N"` heartbeats continue to appear roughly once per minute.
5. Check verdict:
- **VERDICT: APPROVED** with no `P0`, `P1`, or `P2` findings → proceed to Phase 7 (Initialize workspace)
- **VERDICT: APPROVED** with only `P3` findings → optionally fix the `P3` items if they are cheap and safe, then proceed
- **VERDICT: REVISE** or any `P0`, `P1`, or `P2` finding → go to Step 5
- No clear verdict but `P0`, `P1`, and `P2` are all `- None.` → treat as approved
- Helper state `completed-empty-output` → treat as failed review attempt, surface stderr/status, fix invocation or prompt handling, then retry
- Helper state `needs-operator-decision` → surface status log and decide whether to keep waiting, abort, or retry with different helper parameters
- Max rounds (5) reached → proceed with warning
- Helper state `needs-operator-decision` → surface status log and decide whether to extend the timeout, abort, or retry with different helper parameters
- Max rounds (`MAX_ROUNDS`) reached → present the outcome to the user for a manual decision (proceed or stop)
#### Step 5: Revise the Plan
Address each issue the reviewer raised. Update the plan in conversation context and rewrite `/tmp/plan-${REVIEW_ID}.md`.
Address the reviewer findings in priority order (`P0``P1``P2`, then `P3` when practical). Update the plan in conversation context and rewrite `/tmp/plan-${REVIEW_ID}.md`.
Summarize revisions for the user:
@@ -223,7 +327,9 @@ Summarize revisions for the user:
If a revision contradicts the user's explicit requirements, skip it and note it for the user.
#### Step 6: Re-submit to Reviewer (Rounds 2-5)
#### Step 6: Re-submit to Reviewer (Rounds 2-N)
Rewrite `/tmp/plan-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
**If `REVIEWER_CLI` is `codex`:**
@@ -237,8 +343,8 @@ codex exec resume ${CODEX_SESSION_ID} \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE"
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking."
```
If resume fails (session expired), fall back to fresh `codex exec` with context about prior rounds.
@@ -260,8 +366,8 @@ $(cat /tmp/plan-${REVIEW_ID}.md)
Changes made:
[List specific changes]
Re-review the full plan. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review the full plan using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -282,8 +388,8 @@ cursor-agent --resume ${CURSOR_SESSION_ID} -p \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
> /tmp/plan-review-${REVIEW_ID}.json
jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_ID}.md
@@ -302,7 +408,7 @@ Return to Step 4.
**Status:** Approved after N round(s)
[or]
**Status:** Max rounds (5) reached — not fully approved
**Status:** Max rounds (`MAX_ROUNDS`) reached — not fully approved
[Final feedback / remaining concerns]
```
@@ -354,6 +460,27 @@ When handing off to execution, instruct:
Private plan files under `~/.claude/plans/` are planning artifacts and must not be used as execution source of truth.
### Phase 10: Telegram Completion Notification (MANDATORY)
Resolve the Telegram notifier helper from the installed Claude Code skills directory:
```bash
TELEGRAM_NOTIFY_RUNTIME=~/.claude/skills/reviewer-runtime/notify-telegram.sh
```
On every terminal outcome for the create-plan run (approved, max rounds reached, skipped reviewer, or failure), send a Telegram summary if the helper exists and both `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` are configured:
```bash
if [ -x "$TELEGRAM_NOTIFY_RUNTIME" ] && [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
"$TELEGRAM_NOTIFY_RUNTIME" --message "create-plan completed for <plan-folder-name>: <status summary>"
fi
```
Rules:
- Telegram is the only supported completion notification path. Do not use desktop notifications, `say`, email, or any other notifier.
- Notification failures are non-blocking, but they must be surfaced to the user.
- If Telegram is not configured, state that no completion notification was sent.
## Tracker Discipline (MANDATORY)
**ALWAYS update `story-tracker.md` before/after each story. NEVER proceed with stale tracker state.**
@@ -392,6 +519,7 @@ After completing any story:
- [ ] `.gitignore` ignore-rule commit was created if needed
- [ ] Plan directory created under `ai_plan/YYYY-MM-DD-<short-title>/`
- [ ] Reviewer configured or explicitly skipped
- [ ] Max review rounds confirmed (default: 10)
- [ ] Plan review completed (approved or max rounds) — or skipped
- [ ] `original-plan.md` copied from `~/.claude/plans/` plan file
- [ ] `final-transcript.md` present
@@ -399,6 +527,7 @@ After completing any story:
- [ ] `story-tracker.md` created with all stories as `pending`
- [ ] `continuation-runbook.md` present
- [ ] Handoff explicitly says to read runbook first and execute from plan folder
- [ ] Telegram completion notification attempted if configured
## Exit Triggers for Question Phase
User says: "ready", "done", "let's plan", "proceed", "enough questions"

View File

@@ -72,7 +72,10 @@ If the user has already specified a reviewer CLI and model (e.g., "create a plan
- For `cursor`: **run `cursor-agent models` first** to see your account's available models (availability varies by subscription)
- Accept any model string the user provides
Store the chosen `REVIEWER_CLI` and `REVIEWER_MODEL` for Phase 6 (Iterative Plan Review).
3. **Max review rounds for the plan?** (default: 10)
- If the user does not provide a value, set `MAX_ROUNDS=10`.
Store the chosen `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phase 6 (Iterative Plan Review).
### Phase 4: Design (REQUIRED SUB-SKILL)
@@ -84,7 +87,7 @@ Invoke `superpowers:writing-plans`, then break work into milestones and bite-siz
### Phase 6: Iterative Plan Review
Send the plan to the configured reviewer CLI for feedback. Revise and re-submit until approved (max 5 rounds).
Send the plan to the configured reviewer CLI for feedback. Revise and re-submit until approved (default max 10 rounds).
**Skip this phase entirely if reviewer was set to `skip`.**
@@ -109,10 +112,60 @@ Resolve the shared reviewer helper from the installed Codex skills directory:
REVIEWER_RUNTIME=~/.codex/skills/reviewer-runtime/run-review.sh
```
Set helper success-artifact args before writing the command script:
```bash
HELPER_SUCCESS_FILE_ARGS=()
case "$REVIEWER_CLI" in
codex)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/plan-review-${REVIEW_ID}.md)
;;
cursor)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/plan-review-${REVIEW_ID}.json)
;;
esac
```
#### Step 2: Write Plan to Temp File
Write the complete plan (milestones, stories, design decisions, specs) to `/tmp/plan-${REVIEW_ID}.md`.
#### Review Contract (Applies to Every Round)
The reviewer response must use this structure:
```text
## Summary
...
## Findings
### P0
- ...
### P1
- ...
### P2
- ...
### P3
- ...
## Verdict
VERDICT: APPROVED
```
Rules:
- Order findings from `P0` to `P3`.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- Use `- None.` when a severity has no findings.
- `VERDICT: APPROVED` is allowed only when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking.
- The calling agent should still try to fix `P3` findings when they are cheap and safe.
#### Liveness Contract (Applies While Review Is Running)
- The shared reviewer runtime emits `state=in-progress note="In progress N"` heartbeats every 60 seconds while the reviewer child is alive.
- The calling agent must keep waiting as long as a fresh `In progress N` heartbeat keeps arriving roughly once per minute.
- Do not abort just because the review is slow, a soft timeout fired, or a `stall-warning` line appears, as long as the `In progress N` heartbeat continues.
- Treat missing heartbeats, `state=failed`, `state=completed-empty-output`, and `state=needs-operator-decision` as escalation signals.
#### Step 3: Submit to Reviewer (Round 1)
Write the reviewer invocation to `/tmp/plan-review-${REVIEW_ID}.sh` as a bash script:
@@ -136,8 +189,21 @@ codex exec \
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE"
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking."
```
Do not try to capture the Codex session ID yet. When using the helper, extract it from `/tmp/plan-review-${REVIEW_ID}.runner.out` after the command completes (look for `session id: <uuid>`), then store it as `CODEX_SESSION_ID` for resume in subsequent rounds.
@@ -156,8 +222,21 @@ $(cat /tmp/plan-${REVIEW_ID}.md)
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -178,8 +257,21 @@ cursor-agent -p \
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
> /tmp/plan-review-${REVIEW_ID}.json
```
@@ -193,13 +285,16 @@ if [ -x "$REVIEWER_RUNTIME" ]; then
--command-file /tmp/plan-review-${REVIEW_ID}.sh \
--stdout-file /tmp/plan-review-${REVIEW_ID}.runner.out \
--stderr-file /tmp/plan-review-${REVIEW_ID}.stderr \
--status-file /tmp/plan-review-${REVIEW_ID}.status
--status-file /tmp/plan-review-${REVIEW_ID}.status \
"${HELPER_SUCCESS_FILE_ARGS[@]}"
else
echo "Warning: reviewer runtime helper not found at $REVIEWER_RUNTIME; falling back to direct synchronous review." >&2
bash /tmp/plan-review-${REVIEW_ID}.sh >/tmp/plan-review-${REVIEW_ID}.runner.out 2>/tmp/plan-review-${REVIEW_ID}.stderr
fi
```
Run the helper in the foreground and watch its live stdout for `state=in-progress` heartbeats. If your agent environment buffers command output until exit, start the helper in the background and poll `/tmp/plan-review-${REVIEW_ID}.status` separately instead of treating heartbeats as post-hoc-only data.
After the command completes:
- If `REVIEWER_CLI=cursor`, extract the final review text:
@@ -209,6 +304,11 @@ jq -r '.result' /tmp/plan-review-${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-${REVIEW_ID}.md` before verdict parsing.
- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
```bash
cp /tmp/plan-review-${REVIEW_ID}.runner.out /tmp/plan-review-${REVIEW_ID}.md
```
Fallback is allowed only when the helper is missing or not executable.
@@ -227,17 +327,19 @@ Fallback is allowed only when the helper is missing or not executable.
[Reviewer feedback]
```
3. Check verdict:
- **VERDICT: APPROVED** → proceed to Phase 7 (Initialize workspace)
- **VERDICT: REVISE** → go to Step 5
- No clear verdict but positive / no actionable items → treat as approved
4. While the reviewer is still running, keep waiting as long as fresh `state=in-progress note="In progress N"` heartbeats continue to appear roughly once per minute.
5. Check verdict:
- **VERDICT: APPROVED** with no `P0`, `P1`, or `P2` findings → proceed to Phase 7 (Initialize workspace)
- **VERDICT: APPROVED** with only `P3` findings → optionally fix the `P3` items if they are cheap and safe, then proceed
- **VERDICT: REVISE** or any `P0`, `P1`, or `P2` finding → go to Step 5
- No clear verdict but `P0`, `P1`, and `P2` are all `- None.` → treat as approved
- Helper state `completed-empty-output` → treat as failed review attempt, surface stderr/status, fix invocation or prompt handling, then retry
- Helper state `needs-operator-decision` → surface status log and decide whether to keep waiting, abort, or retry with different helper parameters
- Max rounds (5) reached → proceed with warning
- Helper state `needs-operator-decision` → surface status log and decide whether to extend the timeout, abort, or retry with different helper parameters
- Max rounds (`MAX_ROUNDS`) reached → present the outcome to the user for a manual decision (proceed or stop)
#### Step 5: Revise the Plan
Address each issue the reviewer raised. Update the plan in conversation context and rewrite `/tmp/plan-${REVIEW_ID}.md`.
Address the reviewer findings in priority order (`P0``P1``P2`, then `P3` when practical). Update the plan in conversation context and rewrite `/tmp/plan-${REVIEW_ID}.md`.
Summarize revisions for the user:
@@ -248,7 +350,9 @@ Summarize revisions for the user:
If a revision contradicts the user's explicit requirements, skip it and note it for the user.
#### Step 6: Re-submit to Reviewer (Rounds 2-5)
#### Step 6: Re-submit to Reviewer (Rounds 2-N)
Rewrite `/tmp/plan-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
**If `REVIEWER_CLI` is `codex`:**
@@ -262,8 +366,8 @@ codex exec resume ${CODEX_SESSION_ID} \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE"
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking."
```
If resume fails (session expired), fall back to fresh `codex exec` with context about prior rounds.
@@ -285,8 +389,8 @@ $(cat /tmp/plan-${REVIEW_ID}.md)
Changes made:
[List specific changes]
Re-review the full plan. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review the full plan using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -307,8 +411,8 @@ cursor-agent --resume ${CURSOR_SESSION_ID} -p \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
> /tmp/plan-review-${REVIEW_ID}.json
jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_ID}.md
@@ -327,7 +431,7 @@ Return to Step 4.
**Status:** Approved after N round(s)
[or]
**Status:** Max rounds (5) reached — not fully approved
**Status:** Max rounds (`MAX_ROUNDS`) reached — not fully approved
[Final feedback / remaining concerns]
```
@@ -374,19 +478,41 @@ Always instruct the executing agent:
Do not rely on planner-private files during implementation.
### Phase 10: Telegram Completion Notification (MANDATORY)
Resolve the Telegram notifier helper from the installed Codex skills directory:
```bash
TELEGRAM_NOTIFY_RUNTIME=~/.codex/skills/reviewer-runtime/notify-telegram.sh
```
On every terminal outcome for the create-plan run (approved, max rounds reached, skipped reviewer, or failure), send a Telegram summary if the helper exists and both `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` are configured:
```bash
if [ -x "$TELEGRAM_NOTIFY_RUNTIME" ] && [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
"$TELEGRAM_NOTIFY_RUNTIME" --message "create-plan completed for <plan-folder-name>: <status summary>"
fi
```
Rules:
- Telegram is the only supported completion notification path. Do not use desktop notifications, `say`, email, or any other notifier.
- Notification failures are non-blocking, but they must be surfaced to the user.
- If Telegram is not configured, state that no completion notification was sent.
## Quick Reference
| Phase | Action | Required Output |
|---|---|---|
| 1 | Analyze codebase/context | Constraints and known patterns |
| 2 | Gather requirements (one question at a time) | Confirmed scope and success criteria |
| 3 | Configure reviewer CLI and model | `REVIEWER_CLI` and `REVIEWER_MODEL` (or `skip`) |
| 3 | Configure reviewer CLI and model | `REVIEWER_CLI`, `REVIEWER_MODEL`, `MAX_ROUNDS` (or `skip`) |
| 4 | Invoke `superpowers:brainstorming` | Chosen design approach |
| 5 | Invoke `superpowers:writing-plans` | Milestones and bite-sized stories |
| 6 | Iterative plan review (max 5 rounds) | Reviewer approval or max-rounds warning |
| 6 | Iterative plan review (max `MAX_ROUNDS` rounds) | Reviewer approval or max-rounds warning |
| 7 | Initialize `ai_plan/` + `.gitignore` | Local planning workspace ready |
| 8 | Build plan package from templates | Full plan folder with required files |
| 9 | Handoff with runbook-first instruction | Resumable execution context |
| 10 | Send Telegram completion notification | User notified or notification status reported |
## Execution Rules to Include in Plan (MANDATORY)
@@ -413,6 +539,7 @@ Do not rely on planner-private files during implementation.
- Handoff without explicit "read runbook first" direction.
- Skipping the reviewer phase without explicit user opt-out.
- Not capturing the Codex session ID for resume in subsequent review rounds.
- Using any completion notification path other than Telegram.
## Rationalizations and Counters
@@ -440,6 +567,7 @@ Do not rely on planner-private files during implementation.
- [ ] `.gitignore` ignore-rule commit was created if needed
- [ ] Plan directory created under `ai_plan/YYYY-MM-DD-<short-title>/`
- [ ] Reviewer configured or explicitly skipped
- [ ] Max review rounds confirmed (default: 10)
- [ ] Plan review completed (approved or max rounds) — or skipped
- [ ] `original-plan.md` present
- [ ] `final-transcript.md` present
@@ -447,3 +575,4 @@ Do not rely on planner-private files during implementation.
- [ ] `story-tracker.md` present
- [ ] `continuation-runbook.md` present
- [ ] Handoff explicitly says to read runbook first and execute from plan folder
- [ ] Telegram completion notification attempted if configured

View File

@@ -73,7 +73,10 @@ If the user has already specified a reviewer CLI and model (e.g., "create a plan
- For `cursor`: **run `cursor-agent models` first** to see your account's available models (availability varies by subscription)
- Accept any model string the user provides
Store the chosen `REVIEWER_CLI` and `REVIEWER_MODEL` for Phase 6 (Iterative Plan Review).
3. **Max review rounds for the plan?** (default: 10)
- If the user does not provide a value, set `MAX_ROUNDS=10`.
Store the chosen `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phase 6 (Iterative Plan Review).
### Phase 4: Design (REQUIRED SUB-SKILL)
@@ -86,7 +89,7 @@ Story IDs: `S-{milestone}{sequence}`.
### Phase 6: Iterative Plan Review
Send the plan to the configured reviewer CLI for feedback. Revise and re-submit until approved (max 5 rounds).
Send the plan to the configured reviewer CLI for feedback. Revise and re-submit until approved (default max 10 rounds).
**Skip this phase entirely if reviewer was set to `skip`.**
@@ -115,10 +118,60 @@ else
fi
```
Set helper success-artifact args before writing the command script:
```bash
HELPER_SUCCESS_FILE_ARGS=()
case "$REVIEWER_CLI" in
codex)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/plan-review-${REVIEW_ID}.md)
;;
cursor)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/plan-review-${REVIEW_ID}.json)
;;
esac
```
#### Step 2: Write Plan to Temp File
Write the complete plan (milestones, stories, design decisions, specs) to `/tmp/plan-${REVIEW_ID}.md`.
#### Review Contract (Applies to Every Round)
The reviewer response must use this structure:
```text
## Summary
...
## Findings
### P0
- ...
### P1
- ...
### P2
- ...
### P3
- ...
## Verdict
VERDICT: APPROVED
```
Rules:
- Order findings from `P0` to `P3`.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- Use `- None.` when a severity has no findings.
- `VERDICT: APPROVED` is allowed only when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking.
- The calling agent should still try to fix `P3` findings when they are cheap and safe.
#### Liveness Contract (Applies While Review Is Running)
- The shared reviewer runtime emits `state=in-progress note="In progress N"` heartbeats every 60 seconds while the reviewer child is alive.
- The calling agent must keep waiting as long as a fresh `In progress N` heartbeat keeps arriving roughly once per minute.
- Do not abort just because the review is slow, a soft timeout fired, or a `stall-warning` line appears, as long as the `In progress N` heartbeat continues.
- Treat missing heartbeats, `state=failed`, `state=completed-empty-output`, and `state=needs-operator-decision` as escalation signals.
#### Step 3: Submit to Reviewer (Round 1)
Write the reviewer invocation to `/tmp/plan-review-${REVIEW_ID}.sh` as a bash script:
@@ -142,8 +195,21 @@ codex exec \
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE"
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking."
```
Do not try to capture the Codex session ID yet. When using the helper, extract it from `/tmp/plan-review-${REVIEW_ID}.runner.out` after the command completes (look for `session id: <uuid>`), then store it as `CODEX_SESSION_ID` for resume in subsequent rounds.
@@ -162,8 +228,21 @@ $(cat /tmp/plan-${REVIEW_ID}.md)
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -184,8 +263,21 @@ cursor-agent -p \
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
> /tmp/plan-review-${REVIEW_ID}.json
```
@@ -205,13 +297,16 @@ if [ -x "$REVIEWER_RUNTIME" ]; then
--command-file /tmp/plan-review-${REVIEW_ID}.sh \
--stdout-file /tmp/plan-review-${REVIEW_ID}.runner.out \
--stderr-file /tmp/plan-review-${REVIEW_ID}.stderr \
--status-file /tmp/plan-review-${REVIEW_ID}.status
--status-file /tmp/plan-review-${REVIEW_ID}.status \
"${HELPER_SUCCESS_FILE_ARGS[@]}"
else
echo "Warning: reviewer runtime helper not found at $REVIEWER_RUNTIME; falling back to direct synchronous review." >&2
bash /tmp/plan-review-${REVIEW_ID}.sh >/tmp/plan-review-${REVIEW_ID}.runner.out 2>/tmp/plan-review-${REVIEW_ID}.stderr
fi
```
Run the helper in the foreground and watch its live stdout for `state=in-progress` heartbeats. If your agent environment buffers command output until exit, start the helper in the background and poll `/tmp/plan-review-${REVIEW_ID}.status` separately instead of treating heartbeats as post-hoc-only data.
After the command completes:
- If `REVIEWER_CLI=cursor`, extract the final review text:
@@ -221,6 +316,11 @@ jq -r '.result' /tmp/plan-review-${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-${REVIEW_ID}.md` before verdict parsing.
- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
```bash
cp /tmp/plan-review-${REVIEW_ID}.runner.out /tmp/plan-review-${REVIEW_ID}.md
```
#### Step 4: Read Review & Check Verdict
@@ -237,17 +337,19 @@ jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_I
[Reviewer feedback]
```
3. Check verdict:
- **VERDICT: APPROVED** → proceed to Phase 7 (Initialize workspace)
- **VERDICT: REVISE** → go to Step 5
- No clear verdict but positive / no actionable items → treat as approved
4. While the reviewer is still running, keep waiting as long as fresh `state=in-progress note="In progress N"` heartbeats continue to appear roughly once per minute.
5. Check verdict:
- **VERDICT: APPROVED** with no `P0`, `P1`, or `P2` findings → proceed to Phase 7 (Initialize workspace)
- **VERDICT: APPROVED** with only `P3` findings → optionally fix the `P3` items if they are cheap and safe, then proceed
- **VERDICT: REVISE** or any `P0`, `P1`, or `P2` finding → go to Step 5
- No clear verdict but `P0`, `P1`, and `P2` are all `- None.` → treat as approved
- Helper state `completed-empty-output` → treat as failed review attempt, surface stderr/status, fix invocation or prompt handling, then retry
- Helper state `needs-operator-decision` → surface status log and decide whether to keep waiting, abort, or retry with different helper parameters
- Max rounds (5) reached → proceed with warning
- Helper state `needs-operator-decision` → surface status log and decide whether to extend the timeout, abort, or retry with different helper parameters
- Max rounds (`MAX_ROUNDS`) reached → present the outcome to the user for a manual decision (proceed or stop)
#### Step 5: Revise the Plan
Address each issue the reviewer raised. Update the plan in conversation context and rewrite `/tmp/plan-${REVIEW_ID}.md`.
Address the reviewer findings in priority order (`P0``P1``P2`, then `P3` when practical). Update the plan in conversation context and rewrite `/tmp/plan-${REVIEW_ID}.md`.
Summarize revisions for the user:
@@ -258,7 +360,9 @@ Summarize revisions for the user:
If a revision contradicts the user's explicit requirements, skip it and note it for the user.
#### Step 6: Re-submit to Reviewer (Rounds 2-5)
#### Step 6: Re-submit to Reviewer (Rounds 2-N)
Rewrite `/tmp/plan-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
**If `REVIEWER_CLI` is `codex`:**
@@ -272,8 +376,8 @@ codex exec resume ${CODEX_SESSION_ID} \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE"
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking."
```
If resume fails (session expired), fall back to fresh `codex exec` with context about prior rounds.
@@ -295,8 +399,8 @@ $(cat /tmp/plan-${REVIEW_ID}.md)
Changes made:
[List specific changes]
Re-review the full plan. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review the full plan using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -317,8 +421,8 @@ cursor-agent --resume ${CURSOR_SESSION_ID} -p \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
> /tmp/plan-review-${REVIEW_ID}.json
jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_ID}.md
@@ -337,7 +441,7 @@ Return to Step 4.
**Status:** Approved after N round(s)
[or]
**Status:** Max rounds (5) reached — not fully approved
**Status:** Max rounds (`MAX_ROUNDS`) reached — not fully approved
[Final feedback / remaining concerns]
```
@@ -384,19 +488,45 @@ Always instruct the executing agent:
Do not rely on planner-private files during implementation.
### Phase 10: Telegram Completion Notification (MANDATORY)
Resolve the Telegram notifier helper from Cursor's installed skills directory:
```bash
if [ -x .cursor/skills/reviewer-runtime/notify-telegram.sh ]; then
TELEGRAM_NOTIFY_RUNTIME=.cursor/skills/reviewer-runtime/notify-telegram.sh
else
TELEGRAM_NOTIFY_RUNTIME=~/.cursor/skills/reviewer-runtime/notify-telegram.sh
fi
```
On every terminal outcome for the create-plan run (approved, max rounds reached, skipped reviewer, or failure), send a Telegram summary if the helper exists and both `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` are configured:
```bash
if [ -x "$TELEGRAM_NOTIFY_RUNTIME" ] && [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
"$TELEGRAM_NOTIFY_RUNTIME" --message "create-plan completed for <plan-folder-name>: <status summary>"
fi
```
Rules:
- Telegram is the only supported completion notification path. Do not use desktop notifications, `say`, email, or any other notifier.
- Notification failures are non-blocking, but they must be surfaced to the user.
- If Telegram is not configured, state that no completion notification was sent.
## Quick Reference
| Phase | Action | Required Output |
|---|---|---|
| 1 | Analyze codebase/context | Constraints and known patterns |
| 2 | Gather requirements (one question at a time) | Confirmed scope and success criteria |
| 3 | Configure reviewer CLI and model | `REVIEWER_CLI` and `REVIEWER_MODEL` (or `skip`) |
| 3 | Configure reviewer CLI and model | `REVIEWER_CLI`, `REVIEWER_MODEL`, `MAX_ROUNDS` (or `skip`) |
| 4 | Invoke `superpowers:brainstorming` | Chosen design approach |
| 5 | Invoke `superpowers:writing-plans` | Milestones and bite-sized stories |
| 6 | Iterative plan review (max 5 rounds) | Reviewer approval or max-rounds warning |
| 6 | Iterative plan review (max `MAX_ROUNDS` rounds) | Reviewer approval or max-rounds warning |
| 7 | Initialize `ai_plan/` + `.gitignore` | Local planning workspace ready |
| 8 | Build plan package from templates | Full plan folder with required files |
| 9 | Handoff with runbook-first instruction | Resumable execution context |
| 10 | Send Telegram completion notification | User notified or notification status reported |
## Tracker Discipline (MANDATORY)
@@ -437,6 +567,7 @@ After completing any story:
- Omitting one or more required files in the plan package.
- Handoff without explicit "read runbook first" direction.
- Skipping the reviewer phase without explicit user opt-out.
- Using any completion notification path other than Telegram.
## Red Flags - Stop and Correct
@@ -454,6 +585,7 @@ After completing any story:
- [ ] `.gitignore` ignore-rule commit was created if needed
- [ ] Plan directory created under `ai_plan/YYYY-MM-DD-<short-title>/`
- [ ] Reviewer configured or explicitly skipped
- [ ] Max review rounds confirmed (default: 10)
- [ ] Plan review completed (approved or max rounds) — or skipped
- [ ] `original-plan.md` present
- [ ] `final-transcript.md` present
@@ -461,6 +593,7 @@ After completing any story:
- [ ] `story-tracker.md` present
- [ ] `continuation-runbook.md` present
- [ ] Handoff explicitly says to read runbook first and execute from plan folder
- [ ] Telegram completion notification attempted if configured
## Exit Triggers for Question Phase
User says: "ready", "done", "let's plan", "proceed", "enough questions"

View File

@@ -59,7 +59,10 @@ If the user has already specified a reviewer CLI and model (e.g., "create a plan
- For `cursor`: **run `cursor-agent models` first** to see your account's available models (availability varies by subscription)
- Accept any model string the user provides
Store the chosen `REVIEWER_CLI` and `REVIEWER_MODEL` for Phase 7 (Iterative Plan Review).
3. **Max review rounds for the plan?** (default: 10)
- If the user does not provide a value, set `MAX_ROUNDS=10`.
Store the chosen `REVIEWER_CLI`, `REVIEWER_MODEL`, and `MAX_ROUNDS` for Phase 7 (Iterative Plan Review).
### Phase 5: Design (REQUIRED SUB-SKILL)
@@ -78,7 +81,7 @@ Story IDs: `S-{milestone}{sequence}`.
### Phase 7: Iterative Plan Review
Send the plan to the configured reviewer CLI for feedback. Revise and re-submit until approved (max 5 rounds).
Send the plan to the configured reviewer CLI for feedback. Revise and re-submit until approved (default max 10 rounds).
**Skip this phase entirely if reviewer was set to `skip`.**
@@ -103,10 +106,60 @@ Resolve the shared reviewer helper from the installed OpenCode skills directory:
REVIEWER_RUNTIME=~/.config/opencode/skills/reviewer-runtime/run-review.sh
```
Set helper success-artifact args before writing the command script:
```bash
HELPER_SUCCESS_FILE_ARGS=()
case "$REVIEWER_CLI" in
codex)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/plan-review-${REVIEW_ID}.md)
;;
cursor)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/plan-review-${REVIEW_ID}.json)
;;
esac
```
#### Step 2: Write Plan to Temp File
Write the complete plan (milestones, stories, design decisions, specs) to `/tmp/plan-${REVIEW_ID}.md`.
#### Review Contract (Applies to Every Round)
The reviewer response must use this structure:
```text
## Summary
...
## Findings
### P0
- ...
### P1
- ...
### P2
- ...
### P3
- ...
## Verdict
VERDICT: APPROVED
```
Rules:
- Order findings from `P0` to `P3`.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- Use `- None.` when a severity has no findings.
- `VERDICT: APPROVED` is allowed only when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking.
- The calling agent should still try to fix `P3` findings when they are cheap and safe.
#### Liveness Contract (Applies While Review Is Running)
- The shared reviewer runtime emits `state=in-progress note="In progress N"` heartbeats every 60 seconds while the reviewer child is alive.
- The calling agent must keep waiting as long as a fresh `In progress N` heartbeat keeps arriving roughly once per minute.
- Do not abort just because the review is slow, a soft timeout fired, or a `stall-warning` line appears, as long as the `In progress N` heartbeat continues.
- Treat missing heartbeats, `state=failed`, `state=completed-empty-output`, and `state=needs-operator-decision` as escalation signals.
#### Step 3: Submit to Reviewer (Round 1)
Write the reviewer invocation to `/tmp/plan-review-${REVIEW_ID}.sh` as a bash script:
@@ -130,8 +183,21 @@ codex exec \
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE"
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking."
```
Do not try to capture the Codex session ID yet. When using the helper, extract it from `/tmp/plan-review-${REVIEW_ID}.runner.out` after the command completes (look for `session id: <uuid>`), then store it as `CODEX_SESSION_ID` for resume in subsequent rounds.
@@ -150,8 +216,21 @@ $(cat /tmp/plan-${REVIEW_ID}.md)
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -172,8 +251,21 @@ cursor-agent -p \
4. Alternatives — Is there a simpler or better approach?
5. Security — Any security concerns?
Be specific and actionable. If the plan is solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
> /tmp/plan-review-${REVIEW_ID}.json
```
@@ -187,13 +279,16 @@ if [ -x "$REVIEWER_RUNTIME" ]; then
--command-file /tmp/plan-review-${REVIEW_ID}.sh \
--stdout-file /tmp/plan-review-${REVIEW_ID}.runner.out \
--stderr-file /tmp/plan-review-${REVIEW_ID}.stderr \
--status-file /tmp/plan-review-${REVIEW_ID}.status
--status-file /tmp/plan-review-${REVIEW_ID}.status \
"${HELPER_SUCCESS_FILE_ARGS[@]}"
else
echo "Warning: reviewer runtime helper not found at $REVIEWER_RUNTIME; falling back to direct synchronous review." >&2
bash /tmp/plan-review-${REVIEW_ID}.sh >/tmp/plan-review-${REVIEW_ID}.runner.out 2>/tmp/plan-review-${REVIEW_ID}.stderr
fi
```
Run the helper in the foreground and watch its live stdout for `state=in-progress` heartbeats. If your agent environment buffers command output until exit, start the helper in the background and poll `/tmp/plan-review-${REVIEW_ID}.status` separately instead of treating heartbeats as post-hoc-only data.
After the command completes:
- If `REVIEWER_CLI=cursor`, extract the final review text:
@@ -203,6 +298,11 @@ jq -r '.result' /tmp/plan-review-${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-${REVIEW_ID}.md` before verdict parsing.
- If `REVIEWER_CLI=claude`, promote stdout captured by the helper or fallback runner into the markdown review file:
```bash
cp /tmp/plan-review-${REVIEW_ID}.runner.out /tmp/plan-review-${REVIEW_ID}.md
```
#### Step 4: Read Review & Check Verdict
@@ -219,17 +319,19 @@ jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_I
[Reviewer feedback]
```
3. Check verdict:
- **VERDICT: APPROVED** → proceed to Phase 8 (Initialize workspace)
- **VERDICT: REVISE** → go to Step 5
- No clear verdict but positive / no actionable items → treat as approved
4. While the reviewer is still running, keep waiting as long as fresh `state=in-progress note="In progress N"` heartbeats continue to appear roughly once per minute.
5. Check verdict:
- **VERDICT: APPROVED** with no `P0`, `P1`, or `P2` findings → proceed to Phase 8 (Initialize workspace)
- **VERDICT: APPROVED** with only `P3` findings → optionally fix the `P3` items if they are cheap and safe, then proceed
- **VERDICT: REVISE** or any `P0`, `P1`, or `P2` finding → go to Step 5
- No clear verdict but `P0`, `P1`, and `P2` are all `- None.` → treat as approved
- Helper state `completed-empty-output` → treat as failed review attempt, surface stderr/status, fix invocation or prompt handling, then retry
- Helper state `needs-operator-decision` → surface status log and decide whether to keep waiting, abort, or retry with different helper parameters
- Max rounds (5) reached → proceed with warning
- Helper state `needs-operator-decision` → surface status log and decide whether to extend the timeout, abort, or retry with different helper parameters
- Max rounds (`MAX_ROUNDS`) reached → present the outcome to the user for a manual decision (proceed or stop)
#### Step 5: Revise the Plan
Address each issue the reviewer raised. Update the plan in conversation context and rewrite `/tmp/plan-${REVIEW_ID}.md`.
Address the reviewer findings in priority order (`P0``P1``P2`, then `P3` when practical). Update the plan in conversation context and rewrite `/tmp/plan-${REVIEW_ID}.md`.
Summarize revisions for the user:
@@ -240,7 +342,9 @@ Summarize revisions for the user:
If a revision contradicts the user's explicit requirements, skip it and note it for the user.
#### Step 6: Re-submit to Reviewer (Rounds 2-5)
#### Step 6: Re-submit to Reviewer (Rounds 2-N)
Rewrite `/tmp/plan-review-${REVIEW_ID}.sh` for the next round. The script should contain the reviewer invocation only; do not run it directly.
**If `REVIEWER_CLI` is `codex`:**
@@ -254,8 +358,8 @@ codex exec resume ${CODEX_SESSION_ID} \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE"
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking."
```
If resume fails (session expired), fall back to fresh `codex exec` with context about prior rounds.
@@ -277,8 +381,8 @@ $(cat /tmp/plan-${REVIEW_ID}.md)
Changes made:
[List specific changes]
Re-review the full plan. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review the full plan using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -299,8 +403,8 @@ cursor-agent --resume ${CURSOR_SESSION_ID} -p \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
> /tmp/plan-review-${REVIEW_ID}.json
jq -r '.result' /tmp/plan-review-${REVIEW_ID}.json > /tmp/plan-review-${REVIEW_ID}.md
@@ -319,7 +423,7 @@ Return to Step 4.
**Status:** Approved after N round(s)
[or]
**Status:** Max rounds (5) reached — not fully approved
**Status:** Max rounds (`MAX_ROUNDS`) reached — not fully approved
[Final feedback / remaining concerns]
```
@@ -364,6 +468,27 @@ Use templates from this skill's `templates/` folder.
Always instruct the executing agent:
> Read `ai_plan/YYYY-MM-DD-<short-title>/continuation-runbook.md` first, then execute from `ai_plan` files only.
### Phase 11: Telegram Completion Notification (MANDATORY)
Resolve the Telegram notifier helper from the installed OpenCode skills directory:
```bash
TELEGRAM_NOTIFY_RUNTIME=~/.config/opencode/skills/reviewer-runtime/notify-telegram.sh
```
On every terminal outcome for the create-plan run (approved, max rounds reached, skipped reviewer, or failure), send a Telegram summary if the helper exists and both `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` are configured:
```bash
if [ -x "$TELEGRAM_NOTIFY_RUNTIME" ] && [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
"$TELEGRAM_NOTIFY_RUNTIME" --message "create-plan completed for <plan-folder-name>: <status summary>"
fi
```
Rules:
- Telegram is the only supported completion notification path. Do not use desktop notifications, `say`, email, or any other notifier.
- Notification failures are non-blocking, but they must be surfaced to the user.
- If Telegram is not configured, state that no completion notification was sent.
## Tracker Discipline (MANDATORY)
Before starting any story:
@@ -400,6 +525,7 @@ After completing any story:
- [ ] `.gitignore` ignore-rule commit was created if needed
- [ ] Plan directory created under `ai_plan/YYYY-MM-DD-<short-title>/`
- [ ] Reviewer configured or explicitly skipped
- [ ] Max review rounds confirmed (default: 10)
- [ ] Plan review completed (approved or max rounds) — or skipped
- [ ] `original-plan.md` present
- [ ] `final-transcript.md` present
@@ -407,6 +533,7 @@ After completing any story:
- [ ] `story-tracker.md` created with all stories as `pending`
- [ ] `continuation-runbook.md` present
- [ ] Handoff explicitly says to read runbook first and execute from plan folder
- [ ] Telegram completion notification attempted if configured
## Exit Triggers for Question Phase
User says: "ready", "done", "let's plan", "proceed", "enough questions"

View File

@@ -151,6 +151,20 @@ Resolve the shared runtime helper path before writing the command script:
REVIEWER_RUNTIME=~/.claude/skills/reviewer-runtime/run-review.sh
```
Set helper success-artifact args before writing the command script:
```bash
HELPER_SUCCESS_FILE_ARGS=()
case "$REVIEWER_CLI" in
codex)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/milestone-review-${REVIEW_ID}.md)
;;
cursor)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/milestone-review-${REVIEW_ID}.json)
;;
esac
```
#### Step 2: Write Review Payload
Write to `/tmp/milestone-${REVIEW_ID}.md`:
@@ -176,6 +190,42 @@ Write to `/tmp/milestone-${REVIEW_ID}.md`:
[test output with pass/fail counts]
```
#### Review Contract (Applies to Every Round)
The reviewer response must use this structure:
```text
## Summary
...
## Findings
### P0
- ...
### P1
- ...
### P2
- ...
### P3
- ...
## Verdict
VERDICT: APPROVED
```
Rules:
- Order findings from `P0` to `P3`.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- Use `- None.` when a severity has no findings.
- `VERDICT: APPROVED` is allowed only when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking.
- The calling agent should still try to fix `P3` findings when they are cheap and safe.
#### Liveness Contract (Applies While Review Is Running)
- The shared reviewer runtime emits `state=in-progress note="In progress N"` heartbeats every 60 seconds while the reviewer child is alive.
- The calling agent must keep waiting as long as a fresh `In progress N` heartbeat keeps arriving roughly once per minute.
- Do not abort just because the review is slow, a soft timeout fired, or a `stall-warning` line appears, as long as the `In progress N` heartbeat continues.
- Treat missing heartbeats, `state=failed`, `state=completed-empty-output`, and `state=needs-operator-decision` as escalation signals.
#### Step 3: Submit to Reviewer (Round 1)
Write the reviewer invocation to `/tmp/milestone-review-${REVIEW_ID}.sh` as a bash script:
@@ -201,8 +251,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE"
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking."
```
Do not try to capture the Codex session ID yet. When using the helper, extract it from `/tmp/milestone-review-${REVIEW_ID}.runner.out` after the command completes (look for `session id: <uuid>`), then store it as `CODEX_SESSION_ID` for resume in subsequent rounds.
@@ -222,8 +286,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -246,8 +324,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
> /tmp/milestone-review-${REVIEW_ID}.json
```
@@ -261,13 +353,16 @@ if [ -x "$REVIEWER_RUNTIME" ]; then
--command-file /tmp/milestone-review-${REVIEW_ID}.sh \
--stdout-file /tmp/milestone-review-${REVIEW_ID}.runner.out \
--stderr-file /tmp/milestone-review-${REVIEW_ID}.stderr \
--status-file /tmp/milestone-review-${REVIEW_ID}.status
--status-file /tmp/milestone-review-${REVIEW_ID}.status \
"${HELPER_SUCCESS_FILE_ARGS[@]}"
else
echo "Warning: reviewer runtime helper not found at $REVIEWER_RUNTIME; falling back to direct synchronous review." >&2
bash /tmp/milestone-review-${REVIEW_ID}.sh >/tmp/milestone-review-${REVIEW_ID}.runner.out 2>/tmp/milestone-review-${REVIEW_ID}.stderr
fi
```
Run the helper in the foreground and watch its live stdout for `state=in-progress` heartbeats. If your agent environment buffers command output until exit, start the helper in the background and poll `/tmp/milestone-review-${REVIEW_ID}.status` separately instead of treating heartbeats as post-hoc-only data.
After the command completes:
- If `REVIEWER_CLI=cursor`, extract the final review text:
@@ -277,6 +372,11 @@ jq -r '.result' /tmp/milestone-review-${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:
```bash
cp /tmp/milestone-review-${REVIEW_ID}.runner.out /tmp/milestone-review-${REVIEW_ID}.md
```
Fallback is allowed only when the helper is missing or not executable.
@@ -295,17 +395,19 @@ Fallback is allowed only when the helper is missing or not executable.
[Reviewer feedback]
```
4. Check verdict:
- **VERDICT: APPROVED** -> proceed to Phase 4 Step 6 (commit & approve)
- **VERDICT: REVISE** -> go to Step 5
- No clear verdict but positive / no actionable items -> treat as approved
4. While the reviewer is still running, keep waiting as long as fresh `state=in-progress note="In progress N"` heartbeats continue to appear roughly once per minute.
5. Check verdict:
- **VERDICT: APPROVED** with no `P0`, `P1`, or `P2` findings -> proceed to Phase 4 Step 6 (commit & approve)
- **VERDICT: APPROVED** with only `P3` findings -> optionally fix the `P3` items if they are cheap and safe, then proceed
- **VERDICT: REVISE** or any `P0`, `P1`, or `P2` finding -> go to Step 5
- No clear verdict but `P0`, `P1`, and `P2` are all `- None.` -> treat as approved
- Helper state `completed-empty-output` -> treat as failed review attempt, surface stderr/status, fix invocation or prompt handling, then retry
- Helper state `needs-operator-decision` -> surface status log, note any `stall-warning` heartbeat lines as non-terminal operator hints, and decide whether to keep waiting, abort, or retry with different helper parameters
- Helper state `needs-operator-decision` -> surface status log and decide whether to extend the timeout, abort, or retry with different helper parameters
- Max rounds (`MAX_ROUNDS`) reached -> present to user for manual decision (proceed or stop)
#### Step 5: Address Feedback & Re-verify
1. Address each issue the reviewer raised (do NOT commit yet).
1. Address the reviewer findings in priority order (`P0` -> `P1` -> `P2`, then `P3` when practical) (do NOT commit yet).
2. Re-run verification (lint/typecheck/tests) — all must pass.
3. Update `/tmp/milestone-${REVIEW_ID}.md` with new diff and verification output.
@@ -334,8 +436,8 @@ codex exec resume ${CODEX_SESSION_ID} \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE"
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking."
```
If resume fails (session expired), fall back to fresh `codex exec` with context about prior rounds.
@@ -357,12 +459,11 @@ $(cat /tmp/milestone-${REVIEW_ID}.md)
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user \
> /tmp/milestone-review-${REVIEW_ID}.md
--setting-sources user
```
**If `REVIEWER_CLI` is `cursor`:**
@@ -380,8 +481,8 @@ cursor-agent --resume ${CURSOR_SESSION_ID} -p \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
> /tmp/milestone-review-${REVIEW_ID}.json
```
@@ -443,6 +544,27 @@ Present summary:
**Branch:** implement/<plan-folder-name> (merged and deleted)
```
### Phase 8: Telegram Completion Notification (MANDATORY)
Resolve the Telegram notifier helper from the installed Claude Code skills directory:
```bash
TELEGRAM_NOTIFY_RUNTIME=~/.claude/skills/reviewer-runtime/notify-telegram.sh
```
On every terminal outcome for the implement-plan run (fully completed, stopped after max rounds, skipped reviewer, or failure), send a Telegram summary if the helper exists and both `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` are configured:
```bash
if [ -x "$TELEGRAM_NOTIFY_RUNTIME" ] && [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
"$TELEGRAM_NOTIFY_RUNTIME" --message "implement-plan completed for <plan-folder-name>: <status summary>"
fi
```
Rules:
- Telegram is the only supported completion notification path. Do not use desktop notifications, `say`, email, or any other notifier.
- Notification failures are non-blocking, but they must be surfaced to the user.
- If Telegram is not configured, state that no completion notification was sent.
## Tracker Discipline (MANDATORY)
**ALWAYS update `story-tracker.md` before/after each story. NEVER proceed with stale tracker state.**
@@ -477,3 +599,4 @@ Note: Commit hashes are backfilled into story Notes after the milestone commit (
- [ ] Final test suite passes
- [ ] Worktree branch merged to parent and worktree deleted
- [ ] Story tracker updated with final status
- [ ] Telegram completion notification attempted if configured

View File

@@ -184,6 +184,20 @@ Resolve the shared runtime helper path before writing the command script:
REVIEWER_RUNTIME=~/.codex/skills/reviewer-runtime/run-review.sh
```
Set helper success-artifact args before writing the command script:
```bash
HELPER_SUCCESS_FILE_ARGS=()
case "$REVIEWER_CLI" in
codex)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/milestone-review-${REVIEW_ID}.md)
;;
cursor)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/milestone-review-${REVIEW_ID}.json)
;;
esac
```
#### Step 2: Write Review Payload
Write to `/tmp/milestone-${REVIEW_ID}.md`:
@@ -209,6 +223,42 @@ Write to `/tmp/milestone-${REVIEW_ID}.md`:
[test output with pass/fail counts]
```
#### Review Contract (Applies to Every Round)
The reviewer response must use this structure:
```text
## Summary
...
## Findings
### P0
- ...
### P1
- ...
### P2
- ...
### P3
- ...
## Verdict
VERDICT: APPROVED
```
Rules:
- Order findings from `P0` to `P3`.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- Use `- None.` when a severity has no findings.
- `VERDICT: APPROVED` is allowed only when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking.
- The calling agent should still try to fix `P3` findings when they are cheap and safe.
#### Liveness Contract (Applies While Review Is Running)
- The shared reviewer runtime emits `state=in-progress note="In progress N"` heartbeats every 60 seconds while the reviewer child is alive.
- The calling agent must keep waiting as long as a fresh `In progress N` heartbeat keeps arriving roughly once per minute.
- Do not abort just because the review is slow, a soft timeout fired, or a `stall-warning` line appears, as long as the `In progress N` heartbeat continues.
- Treat missing heartbeats, `state=failed`, `state=completed-empty-output`, and `state=needs-operator-decision` as escalation signals.
#### Step 3: Submit to Reviewer (Round 1)
Write the reviewer invocation to `/tmp/milestone-review-${REVIEW_ID}.sh` as a bash script:
@@ -234,8 +284,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE"
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking."
```
Do not try to capture the Codex session ID yet. When using the helper, extract it from `/tmp/milestone-review-${REVIEW_ID}.runner.out` after the command completes (look for `session id: <uuid>`), then store it as `CODEX_SESSION_ID` for resume in subsequent rounds.
@@ -255,8 +319,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -279,8 +357,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
> /tmp/milestone-review-${REVIEW_ID}.json
```
@@ -294,13 +386,16 @@ if [ -x "$REVIEWER_RUNTIME" ]; then
--command-file /tmp/milestone-review-${REVIEW_ID}.sh \
--stdout-file /tmp/milestone-review-${REVIEW_ID}.runner.out \
--stderr-file /tmp/milestone-review-${REVIEW_ID}.stderr \
--status-file /tmp/milestone-review-${REVIEW_ID}.status
--status-file /tmp/milestone-review-${REVIEW_ID}.status \
"${HELPER_SUCCESS_FILE_ARGS[@]}"
else
echo "Warning: reviewer runtime helper not found at $REVIEWER_RUNTIME; falling back to direct synchronous review." >&2
bash /tmp/milestone-review-${REVIEW_ID}.sh >/tmp/milestone-review-${REVIEW_ID}.runner.out 2>/tmp/milestone-review-${REVIEW_ID}.stderr
fi
```
Run the helper in the foreground and watch its live stdout for `state=in-progress` heartbeats. If your agent environment buffers command output until exit, start the helper in the background and poll `/tmp/milestone-review-${REVIEW_ID}.status` separately instead of treating heartbeats as post-hoc-only data.
After the command completes:
- If `REVIEWER_CLI=cursor`, extract the final review text:
@@ -310,6 +405,11 @@ jq -r '.result' /tmp/milestone-review-${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:
```bash
cp /tmp/milestone-review-${REVIEW_ID}.runner.out /tmp/milestone-review-${REVIEW_ID}.md
```
Fallback is allowed only when the helper is missing or not executable.
@@ -328,17 +428,19 @@ Fallback is allowed only when the helper is missing or not executable.
[Reviewer feedback]
```
4. Check verdict:
- **VERDICT: APPROVED** -> proceed to Phase 4 Step 6 (commit & approve)
- **VERDICT: REVISE** -> go to Step 5
- No clear verdict but positive / no actionable items -> treat as approved
4. While the reviewer is still running, keep waiting as long as fresh `state=in-progress note="In progress N"` heartbeats continue to appear roughly once per minute.
5. Check verdict:
- **VERDICT: APPROVED** with no `P0`, `P1`, or `P2` findings -> proceed to Phase 4 Step 6 (commit & approve)
- **VERDICT: APPROVED** with only `P3` findings -> optionally fix the `P3` items if they are cheap and safe, then proceed
- **VERDICT: REVISE** or any `P0`, `P1`, or `P2` finding -> go to Step 5
- No clear verdict but `P0`, `P1`, and `P2` are all `- None.` -> treat as approved
- Helper state `completed-empty-output` -> treat as failed review attempt, surface stderr/status, fix invocation or prompt handling, then retry
- Helper state `needs-operator-decision` -> surface status log, note any `stall-warning` heartbeat lines as non-terminal operator hints, and decide whether to keep waiting, abort, or retry with different helper parameters
- Helper state `needs-operator-decision` -> surface status log and decide whether to extend the timeout, abort, or retry with different helper parameters
- Max rounds (`MAX_ROUNDS`) reached -> present to user for manual decision (proceed or stop)
#### Step 5: Address Feedback & Re-verify
1. Address each issue the reviewer raised (do NOT commit yet).
1. Address the reviewer findings in priority order (`P0` -> `P1` -> `P2`, then `P3` when practical) (do NOT commit yet).
2. Re-run verification (lint/typecheck/tests) — all must pass.
3. Update `/tmp/milestone-${REVIEW_ID}.md` with new diff and verification output.
@@ -367,8 +469,8 @@ codex exec resume ${CODEX_SESSION_ID} \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE"
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking."
```
If resume fails (session expired), fall back to fresh `codex exec` with context about prior rounds.
@@ -390,12 +492,11 @@ $(cat /tmp/milestone-${REVIEW_ID}.md)
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user \
> /tmp/milestone-review-${REVIEW_ID}.md
--setting-sources user
```
**If `REVIEWER_CLI` is `cursor`:**
@@ -413,8 +514,8 @@ cursor-agent --resume ${CURSOR_SESSION_ID} -p \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
> /tmp/milestone-review-${REVIEW_ID}.json
```
@@ -476,6 +577,27 @@ Present summary:
**Branch:** implement/<plan-folder-name> (merged and deleted)
```
### Phase 8: Telegram Completion Notification (MANDATORY)
Resolve the Telegram notifier helper from the installed Codex skills directory:
```bash
TELEGRAM_NOTIFY_RUNTIME=~/.codex/skills/reviewer-runtime/notify-telegram.sh
```
On every terminal outcome for the implement-plan run (fully completed, stopped after max rounds, skipped reviewer, or failure), send a Telegram summary if the helper exists and both `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` are configured:
```bash
if [ -x "$TELEGRAM_NOTIFY_RUNTIME" ] && [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
"$TELEGRAM_NOTIFY_RUNTIME" --message "implement-plan completed for <plan-folder-name>: <status summary>"
fi
```
Rules:
- Telegram is the only supported completion notification path. Do not use desktop notifications, `say`, email, or any other notifier.
- Notification failures are non-blocking, but they must be surfaced to the user.
- If Telegram is not configured, state that no completion notification was sent.
## Quick Reference
| Phase | Action | Required Output |
@@ -487,6 +609,7 @@ Present summary:
| 5 | Milestone review loop (per milestone) | Reviewer approval or max rounds + user override |
| 6 | Invoke `superpowers:finishing-a-development-branch` | Branch merged to parent, worktree deleted |
| 7 | Final report | Summary presented |
| 8 | Send Telegram completion notification | User notified or notification status reported |
## Tracker Discipline (MANDATORY)
@@ -514,6 +637,7 @@ Note: Commit hashes are backfilled into story Notes after the milestone commit (
- Not capturing the Codex session ID for resume in subsequent review rounds.
- Forgetting to update `story-tracker.md` between stories.
- Creating a new worktree when one already exists for a resumed plan.
- Using any completion notification path other than Telegram.
## Rationalizations and Counters
@@ -551,3 +675,4 @@ Note: Commit hashes are backfilled into story Notes after the milestone commit (
- [ ] Final test suite passes
- [ ] Worktree branch merged to parent and worktree deleted
- [ ] Story tracker updated with final status
- [ ] Telegram completion notification attempted if configured

View File

@@ -188,6 +188,20 @@ else
fi
```
Set helper success-artifact args before writing the command script:
```bash
HELPER_SUCCESS_FILE_ARGS=()
case "$REVIEWER_CLI" in
codex)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/milestone-review-${REVIEW_ID}.md)
;;
cursor)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/milestone-review-${REVIEW_ID}.json)
;;
esac
```
#### Step 2: Write Review Payload
Write to `/tmp/milestone-${REVIEW_ID}.md`:
@@ -213,6 +227,42 @@ Write to `/tmp/milestone-${REVIEW_ID}.md`:
[test output with pass/fail counts]
```
#### Review Contract (Applies to Every Round)
The reviewer response must use this structure:
```text
## Summary
...
## Findings
### P0
- ...
### P1
- ...
### P2
- ...
### P3
- ...
## Verdict
VERDICT: APPROVED
```
Rules:
- Order findings from `P0` to `P3`.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- Use `- None.` when a severity has no findings.
- `VERDICT: APPROVED` is allowed only when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking.
- The calling agent should still try to fix `P3` findings when they are cheap and safe.
#### Liveness Contract (Applies While Review Is Running)
- The shared reviewer runtime emits `state=in-progress note="In progress N"` heartbeats every 60 seconds while the reviewer child is alive.
- The calling agent must keep waiting as long as a fresh `In progress N` heartbeat keeps arriving roughly once per minute.
- Do not abort just because the review is slow, a soft timeout fired, or a `stall-warning` line appears, as long as the `In progress N` heartbeat continues.
- Treat missing heartbeats, `state=failed`, `state=completed-empty-output`, and `state=needs-operator-decision` as escalation signals.
#### Step 3: Submit to Reviewer (Round 1)
Write the reviewer invocation to `/tmp/milestone-review-${REVIEW_ID}.sh` as a bash script:
@@ -238,8 +288,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE"
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking."
```
Do not try to capture the Codex session ID yet. When using the helper, extract it from `/tmp/milestone-review-${REVIEW_ID}.runner.out` after the command completes (look for `session id: <uuid>`), then store it as `CODEX_SESSION_ID` for resume in subsequent rounds.
@@ -259,8 +323,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -283,8 +361,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
> /tmp/milestone-review-${REVIEW_ID}.json
```
@@ -304,13 +396,16 @@ if [ -x "$REVIEWER_RUNTIME" ]; then
--command-file /tmp/milestone-review-${REVIEW_ID}.sh \
--stdout-file /tmp/milestone-review-${REVIEW_ID}.runner.out \
--stderr-file /tmp/milestone-review-${REVIEW_ID}.stderr \
--status-file /tmp/milestone-review-${REVIEW_ID}.status
--status-file /tmp/milestone-review-${REVIEW_ID}.status \
"${HELPER_SUCCESS_FILE_ARGS[@]}"
else
echo "Warning: reviewer runtime helper not found at $REVIEWER_RUNTIME; falling back to direct synchronous review." >&2
bash /tmp/milestone-review-${REVIEW_ID}.sh >/tmp/milestone-review-${REVIEW_ID}.runner.out 2>/tmp/milestone-review-${REVIEW_ID}.stderr
fi
```
Run the helper in the foreground and watch its live stdout for `state=in-progress` heartbeats. If your agent environment buffers command output until exit, start the helper in the background and poll `/tmp/milestone-review-${REVIEW_ID}.status` separately instead of treating heartbeats as post-hoc-only data.
After the command completes:
- If `REVIEWER_CLI=cursor`, extract the final review text:
@@ -320,6 +415,11 @@ jq -r '.result' /tmp/milestone-review-${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:
```bash
cp /tmp/milestone-review-${REVIEW_ID}.runner.out /tmp/milestone-review-${REVIEW_ID}.md
```
Fallback is allowed only when the helper is missing or not executable.
@@ -338,17 +438,19 @@ Fallback is allowed only when the helper is missing or not executable.
[Reviewer feedback]
```
4. Check verdict:
- **VERDICT: APPROVED** -> proceed to Phase 4 Step 6 (commit & approve)
- **VERDICT: REVISE** -> go to Step 5
- No clear verdict but positive / no actionable items -> treat as approved
4. While the reviewer is still running, keep waiting as long as fresh `state=in-progress note="In progress N"` heartbeats continue to appear roughly once per minute.
5. Check verdict:
- **VERDICT: APPROVED** with no `P0`, `P1`, or `P2` findings -> proceed to Phase 4 Step 6 (commit & approve)
- **VERDICT: APPROVED** with only `P3` findings -> optionally fix the `P3` items if they are cheap and safe, then proceed
- **VERDICT: REVISE** or any `P0`, `P1`, or `P2` finding -> go to Step 5
- No clear verdict but `P0`, `P1`, and `P2` are all `- None.` -> treat as approved
- Helper state `completed-empty-output` -> treat as failed review attempt, surface stderr/status, fix invocation or prompt handling, then retry
- Helper state `needs-operator-decision` -> surface status log, note any `stall-warning` heartbeat lines as non-terminal operator hints, and decide whether to keep waiting, abort, or retry with different helper parameters
- Helper state `needs-operator-decision` -> surface status log and decide whether to extend the timeout, abort, or retry with different helper parameters
- Max rounds (`MAX_ROUNDS`) reached -> present to user for manual decision (proceed or stop)
#### Step 5: Address Feedback & Re-verify
1. Address each issue the reviewer raised (do NOT commit yet).
1. Address the reviewer findings in priority order (`P0` -> `P1` -> `P2`, then `P3` when practical) (do NOT commit yet).
2. Re-run verification (lint/typecheck/tests) — all must pass.
3. Update `/tmp/milestone-${REVIEW_ID}.md` with new diff and verification output.
@@ -377,8 +479,8 @@ codex exec resume ${CODEX_SESSION_ID} \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE"
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking."
```
If resume fails (session expired), fall back to fresh `codex exec` with context about prior rounds.
@@ -400,12 +502,11 @@ $(cat /tmp/milestone-${REVIEW_ID}.md)
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user \
> /tmp/milestone-review-${REVIEW_ID}.md
--setting-sources user
```
**If `REVIEWER_CLI` is `cursor`:**
@@ -423,8 +524,8 @@ cursor-agent --resume ${CURSOR_SESSION_ID} -p \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
> /tmp/milestone-review-${REVIEW_ID}.json
```
@@ -486,6 +587,31 @@ Present summary:
**Branch:** implement/<plan-folder-name> (merged and deleted)
```
### Phase 8: Telegram Completion Notification (MANDATORY)
Resolve the Telegram notifier helper from Cursor's installed skills directory:
```bash
if [ -x .cursor/skills/reviewer-runtime/notify-telegram.sh ]; then
TELEGRAM_NOTIFY_RUNTIME=.cursor/skills/reviewer-runtime/notify-telegram.sh
else
TELEGRAM_NOTIFY_RUNTIME=~/.cursor/skills/reviewer-runtime/notify-telegram.sh
fi
```
On every terminal outcome for the implement-plan run (fully completed, stopped after max rounds, skipped reviewer, or failure), send a Telegram summary if the helper exists and both `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` are configured:
```bash
if [ -x "$TELEGRAM_NOTIFY_RUNTIME" ] && [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
"$TELEGRAM_NOTIFY_RUNTIME" --message "implement-plan completed for <plan-folder-name>: <status summary>"
fi
```
Rules:
- Telegram is the only supported completion notification path. Do not use desktop notifications, `say`, email, or any other notifier.
- Notification failures are non-blocking, but they must be surfaced to the user.
- If Telegram is not configured, state that no completion notification was sent.
## Quick Reference
| Phase | Action | Required Output |
@@ -497,6 +623,7 @@ Present summary:
| 5 | Milestone review loop (per milestone) | Reviewer approval or max rounds + user override |
| 6 | Invoke `superpowers:finishing-a-development-branch` | Branch merged to parent, worktree deleted |
| 7 | Final report | Summary presented |
| 8 | Send Telegram completion notification | User notified or notification status reported |
## Tracker Discipline (MANDATORY)
@@ -522,6 +649,7 @@ Note: Commit hashes are backfilled into story Notes after the milestone commit (
- Skipping worktree setup and working directly on the main branch.
- Forgetting to update `story-tracker.md` between stories.
- Creating a new worktree when one already exists for a resumed plan.
- Using any completion notification path other than Telegram.
## Red Flags - Stop and Correct
@@ -550,3 +678,4 @@ Note: Commit hashes are backfilled into story Notes after the milestone commit (
- [ ] Final test suite passes
- [ ] Worktree branch merged to parent and worktree deleted
- [ ] Story tracker updated with final status
- [ ] Telegram completion notification attempted if configured

View File

@@ -169,6 +169,20 @@ Resolve the shared runtime helper path before writing the command script:
REVIEWER_RUNTIME=~/.config/opencode/skills/reviewer-runtime/run-review.sh
```
Set helper success-artifact args before writing the command script:
```bash
HELPER_SUCCESS_FILE_ARGS=()
case "$REVIEWER_CLI" in
codex)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/milestone-review-${REVIEW_ID}.md)
;;
cursor)
HELPER_SUCCESS_FILE_ARGS+=(--success-file /tmp/milestone-review-${REVIEW_ID}.json)
;;
esac
```
#### Step 2: Write Review Payload
Write to `/tmp/milestone-${REVIEW_ID}.md`:
@@ -194,6 +208,42 @@ Write to `/tmp/milestone-${REVIEW_ID}.md`:
[test output with pass/fail counts]
```
#### Review Contract (Applies to Every Round)
The reviewer response must use this structure:
```text
## Summary
...
## Findings
### P0
- ...
### P1
- ...
### P2
- ...
### P3
- ...
## Verdict
VERDICT: APPROVED
```
Rules:
- Order findings from `P0` to `P3`.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- Use `- None.` when a severity has no findings.
- `VERDICT: APPROVED` is allowed only when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking.
- The calling agent should still try to fix `P3` findings when they are cheap and safe.
#### Liveness Contract (Applies While Review Is Running)
- The shared reviewer runtime emits `state=in-progress note="In progress N"` heartbeats every 60 seconds while the reviewer child is alive.
- The calling agent must keep waiting as long as a fresh `In progress N` heartbeat keeps arriving roughly once per minute.
- Do not abort just because the review is slow, a soft timeout fired, or a `stall-warning` line appears, as long as the `In progress N` heartbeat continues.
- Treat missing heartbeats, `state=failed`, `state=completed-empty-output`, and `state=needs-operator-decision` as escalation signals.
#### Step 3: Submit to Reviewer (Round 1)
Write the reviewer invocation to `/tmp/milestone-review-${REVIEW_ID}.sh` as a bash script:
@@ -219,8 +269,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE"
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking."
```
Do not try to capture the Codex session ID yet. When using the helper, extract it from `/tmp/milestone-review-${REVIEW_ID}.runner.out` after the command completes (look for `session id: <uuid>`), then store it as `CODEX_SESSION_ID` for resume in subsequent rounds.
@@ -240,8 +304,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user
@@ -264,8 +342,22 @@ Evaluate:
4. Test coverage — Are changes adequately tested?
5. Security — Any security concerns introduced?
Be specific and actionable. If solid, end with exactly: VERDICT: APPROVED
If changes are needed, end with exactly: VERDICT: REVISE" \
Return exactly these sections in order:
## Summary
## Findings
### P0
### P1
### P2
### P3
## Verdict
Rules:
- Order findings from highest severity to lowest.
- Use `- None.` when a severity has no findings.
- `P0` = total blocker, `P1` = major risk, `P2` = must-fix before approval, `P3` = cosmetic / nice to have.
- End with exactly one verdict line: `VERDICT: APPROVED` or `VERDICT: REVISE`
- `VERDICT: APPROVED` is allowed only when there are no `P0`, `P1`, or `P2` findings. `P3` findings are non-blocking." \
> /tmp/milestone-review-${REVIEW_ID}.json
```
@@ -279,13 +371,16 @@ if [ -x "$REVIEWER_RUNTIME" ]; then
--command-file /tmp/milestone-review-${REVIEW_ID}.sh \
--stdout-file /tmp/milestone-review-${REVIEW_ID}.runner.out \
--stderr-file /tmp/milestone-review-${REVIEW_ID}.stderr \
--status-file /tmp/milestone-review-${REVIEW_ID}.status
--status-file /tmp/milestone-review-${REVIEW_ID}.status \
"${HELPER_SUCCESS_FILE_ARGS[@]}"
else
echo "Warning: reviewer runtime helper not found at $REVIEWER_RUNTIME; falling back to direct synchronous review." >&2
bash /tmp/milestone-review-${REVIEW_ID}.sh >/tmp/milestone-review-${REVIEW_ID}.runner.out 2>/tmp/milestone-review-${REVIEW_ID}.stderr
fi
```
Run the helper in the foreground and watch its live stdout for `state=in-progress` heartbeats. If your agent environment buffers command output until exit, start the helper in the background and poll `/tmp/milestone-review-${REVIEW_ID}.status` separately instead of treating heartbeats as post-hoc-only data.
After the command completes:
- If `REVIEWER_CLI=cursor`, extract the final review text:
@@ -295,6 +390,11 @@ jq -r '.result' /tmp/milestone-review-${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:
```bash
cp /tmp/milestone-review-${REVIEW_ID}.runner.out /tmp/milestone-review-${REVIEW_ID}.md
```
Fallback is allowed only when the helper is missing or not executable.
@@ -313,17 +413,19 @@ Fallback is allowed only when the helper is missing or not executable.
[Reviewer feedback]
```
4. Check verdict:
- **VERDICT: APPROVED** -> proceed to Phase 5 Step 6 (commit & approve)
- **VERDICT: REVISE** -> go to Step 5
- No clear verdict but positive / no actionable items -> treat as approved
4. While the reviewer is still running, keep waiting as long as fresh `state=in-progress note="In progress N"` heartbeats continue to appear roughly once per minute.
5. Check verdict:
- **VERDICT: APPROVED** with no `P0`, `P1`, or `P2` findings -> proceed to Phase 5 Step 6 (commit & approve)
- **VERDICT: APPROVED** with only `P3` findings -> optionally fix the `P3` items if they are cheap and safe, then proceed
- **VERDICT: REVISE** or any `P0`, `P1`, or `P2` finding -> go to Step 5
- No clear verdict but `P0`, `P1`, and `P2` are all `- None.` -> treat as approved
- Helper state `completed-empty-output` -> treat as failed review attempt, surface stderr/status, fix invocation or prompt handling, then retry
- Helper state `needs-operator-decision` -> surface status log, note any `stall-warning` heartbeat lines as non-terminal operator hints, and decide whether to keep waiting, abort, or retry with different helper parameters
- Helper state `needs-operator-decision` -> surface status log and decide whether to extend the timeout, abort, or retry with different helper parameters
- Max rounds (`MAX_ROUNDS`) reached -> present to user for manual decision (proceed or stop)
#### Step 5: Address Feedback & Re-verify
1. Address each issue the reviewer raised (do NOT commit yet).
1. Address the reviewer findings in priority order (`P0` -> `P1` -> `P2`, then `P3` when practical) (do NOT commit yet).
2. Re-run verification (lint/typecheck/tests) — all must pass.
3. Update `/tmp/milestone-${REVIEW_ID}.md` with new diff and verification output.
@@ -352,8 +454,8 @@ codex exec resume ${CODEX_SESSION_ID} \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE"
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking."
```
If resume fails (session expired), fall back to fresh `codex exec` with context about prior rounds.
@@ -375,12 +477,11 @@ $(cat /tmp/milestone-${REVIEW_ID}.md)
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
--model ${REVIEWER_MODEL} \
--strict-mcp-config \
--setting-sources user \
> /tmp/milestone-review-${REVIEW_ID}.md
--setting-sources user
```
**If `REVIEWER_CLI` is `cursor`:**
@@ -398,8 +499,8 @@ cursor-agent --resume ${CURSOR_SESSION_ID} -p \
Changes made:
[List specific changes]
Re-review. If solid, end with: VERDICT: APPROVED
If more changes needed, end with: VERDICT: REVISE" \
Re-review using the same `## Summary`, `## Findings`, and `## Verdict` structure as before.
Keep findings ordered `P0` to `P3`, use `- None.` when a severity has no findings, and only use `VERDICT: APPROVED` when no `P0`, `P1`, or `P2` findings remain. `P3` findings are non-blocking." \
> /tmp/milestone-review-${REVIEW_ID}.json
```
@@ -461,6 +562,27 @@ Present summary:
**Branch:** implement/<plan-folder-name> (merged and deleted)
```
### Phase 9: Telegram Completion Notification (MANDATORY)
Resolve the Telegram notifier helper from the installed OpenCode skills directory:
```bash
TELEGRAM_NOTIFY_RUNTIME=~/.config/opencode/skills/reviewer-runtime/notify-telegram.sh
```
On every terminal outcome for the implement-plan run (fully completed, stopped after max rounds, skipped reviewer, or failure), send a Telegram summary if the helper exists and both `TELEGRAM_BOT_TOKEN` and `TELEGRAM_CHAT_ID` are configured:
```bash
if [ -x "$TELEGRAM_NOTIFY_RUNTIME" ] && [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
"$TELEGRAM_NOTIFY_RUNTIME" --message "implement-plan completed for <plan-folder-name>: <status summary>"
fi
```
Rules:
- Telegram is the only supported completion notification path. Do not use desktop notifications, `say`, email, or any other notifier.
- Notification failures are non-blocking, but they must be surfaced to the user.
- If Telegram is not configured, state that no completion notification was sent.
## Tracker Discipline (MANDATORY)
Before starting any story:
@@ -493,3 +615,4 @@ Note: Commit hashes are backfilled into story Notes after the milestone commit (
- [ ] Final test suite passes
- [ ] Worktree branch merged to parent and worktree deleted
- [ ] Story tracker updated with final status
- [ ] Telegram completion notification attempted if configured

View File

@@ -0,0 +1,99 @@
#!/usr/bin/env bash
set -euo pipefail
DEFAULT_API_BASE_URL="https://api.telegram.org"
DEFAULT_PARSE_MODE="HTML"
MAX_MESSAGE_LENGTH=4096
BOT_TOKEN=${TELEGRAM_BOT_TOKEN:-}
CHAT_ID=${TELEGRAM_CHAT_ID:-}
API_BASE_URL=${TELEGRAM_API_BASE_URL:-$DEFAULT_API_BASE_URL}
PARSE_MODE=${TELEGRAM_PARSE_MODE:-$DEFAULT_PARSE_MODE}
MESSAGE=""
MESSAGE_FILE=""
usage() {
cat <<'EOF'
Usage:
notify-telegram.sh --message <text> [--bot-token <token>] [--chat-id <id>] [--api-base-url <url>]
notify-telegram.sh --message-file <path> [--bot-token <token>] [--chat-id <id>] [--api-base-url <url>]
Environment fallbacks:
TELEGRAM_BOT_TOKEN
TELEGRAM_CHAT_ID
TELEGRAM_API_BASE_URL
TELEGRAM_PARSE_MODE
EOF
}
fail_usage() {
echo "Error: $*" >&2
usage >&2
exit 2
}
parse_args() {
while [[ $# -gt 0 ]]; do
case "$1" in
--bot-token)
BOT_TOKEN=${2:-}
shift 2
;;
--chat-id)
CHAT_ID=${2:-}
shift 2
;;
--api-base-url)
API_BASE_URL=${2:-}
shift 2
;;
--message)
MESSAGE=${2:-}
shift 2
;;
--message-file)
MESSAGE_FILE=${2:-}
shift 2
;;
--help|-h)
usage
exit 0
;;
*)
fail_usage "unknown argument: $1"
;;
esac
done
if [[ -n "$MESSAGE" && -n "$MESSAGE_FILE" ]]; then
fail_usage "use either --message or --message-file, not both"
fi
if [[ -n "$MESSAGE_FILE" ]]; then
[[ -r "$MESSAGE_FILE" ]] || fail_usage "message file is not readable: $MESSAGE_FILE"
MESSAGE=$(<"$MESSAGE_FILE")
fi
[[ -n "$MESSAGE" ]] || fail_usage "message is required"
[[ -n "$BOT_TOKEN" ]] || fail_usage "bot token is required (use --bot-token or TELEGRAM_BOT_TOKEN)"
[[ -n "$CHAT_ID" ]] || fail_usage "chat id is required (use --chat-id or TELEGRAM_CHAT_ID)"
command -v curl >/dev/null 2>&1 || fail_usage "curl is required"
if [[ ${#MESSAGE} -gt "$MAX_MESSAGE_LENGTH" ]]; then
MESSAGE=${MESSAGE:0:$MAX_MESSAGE_LENGTH}
fi
}
main() {
parse_args "$@"
curl -fsS -X POST \
"${API_BASE_URL%/}/bot${BOT_TOKEN}/sendMessage" \
--data-urlencode "chat_id=${CHAT_ID}" \
--data-urlencode "text=${MESSAGE}" \
--data-urlencode "parse_mode=${PARSE_MODE}" \
--data-urlencode "disable_web_page_preview=true" \
>/dev/null
}
main "$@"

View File

@@ -2,6 +2,7 @@
set -euo pipefail
DEFAULT_POLL_SECONDS=10
DEFAULT_HEARTBEAT_SECONDS=60
DEFAULT_SOFT_TIMEOUT_SECONDS=600
DEFAULT_STALL_WARNING_SECONDS=300
DEFAULT_HARD_TIMEOUT_SECONDS=1800
@@ -12,7 +13,9 @@ COMMAND_FILE=""
STDOUT_FILE=""
STDERR_FILE=""
STATUS_FILE=""
SUCCESS_FILES=()
POLL_SECONDS=$DEFAULT_POLL_SECONDS
HEARTBEAT_SECONDS=$DEFAULT_HEARTBEAT_SECONDS
SOFT_TIMEOUT_SECONDS=$DEFAULT_SOFT_TIMEOUT_SECONDS
STALL_WARNING_SECONDS=$DEFAULT_STALL_WARNING_SECONDS
HARD_TIMEOUT_SECONDS=$DEFAULT_HARD_TIMEOUT_SECONDS
@@ -29,7 +32,9 @@ Usage:
--stdout-file <path> \
--stderr-file <path> \
--status-file <path> \
[--success-file <path>] \
[--poll-seconds <int>] \
[--heartbeat-seconds <int>] \
[--soft-timeout-seconds <int>] \
[--stall-warning-seconds <int>] \
[--hard-timeout-seconds <int>]
@@ -55,6 +60,25 @@ escape_note() {
printf '%s' "$note"
}
join_success_files() {
if [[ ${#SUCCESS_FILES[@]} -eq 0 ]]; then
printf ''
return 0
fi
local joined=""
local path
for path in "${SUCCESS_FILES[@]}"; do
if [[ -n "$joined" ]]; then
joined+=", "
fi
joined+="$path"
done
printf '%s' "$joined"
}
iso_timestamp() {
date -u +"%Y-%m-%dT%H:%M:%SZ"
}
@@ -147,10 +171,18 @@ parse_args() {
STATUS_FILE=${2:-}
shift 2
;;
--success-file)
SUCCESS_FILES+=("${2:-}")
shift 2
;;
--poll-seconds)
POLL_SECONDS=${2:-}
shift 2
;;
--heartbeat-seconds)
HEARTBEAT_SECONDS=${2:-}
shift 2
;;
--soft-timeout-seconds)
SOFT_TIMEOUT_SECONDS=${2:-}
shift 2
@@ -179,11 +211,13 @@ parse_args() {
[[ -n "$STATUS_FILE" ]] || fail_usage "--status-file is required"
require_integer "poll-seconds" "$POLL_SECONDS"
require_integer "heartbeat-seconds" "$HEARTBEAT_SECONDS"
require_integer "soft-timeout-seconds" "$SOFT_TIMEOUT_SECONDS"
require_integer "stall-warning-seconds" "$STALL_WARNING_SECONDS"
require_integer "hard-timeout-seconds" "$HARD_TIMEOUT_SECONDS"
[[ "$POLL_SECONDS" -gt 0 ]] || fail_usage "poll-seconds must be > 0"
[[ "$HEARTBEAT_SECONDS" -gt 0 ]] || fail_usage "heartbeat-seconds must be > 0"
[[ "$SOFT_TIMEOUT_SECONDS" -gt 0 ]] || fail_usage "soft-timeout-seconds must be > 0"
[[ "$STALL_WARNING_SECONDS" -gt 0 ]] || fail_usage "stall-warning-seconds must be > 0"
[[ "$HARD_TIMEOUT_SECONDS" -gt 0 ]] || fail_usage "hard-timeout-seconds must be > 0"
@@ -227,8 +261,10 @@ main() {
local last_stdout_bytes=0
local last_stderr_bytes=0
local last_output_change_time=$START_TIME
local last_heartbeat_time=$START_TIME
local soft_timeout_logged=0
local stall_warning_logged=0
local heartbeat_count=0
while kill -0 "$CHILD_PID" 2>/dev/null; do
sleep "$POLL_SECONDS"
@@ -239,6 +275,12 @@ main() {
stdout_bytes=$(file_bytes "$STDOUT_FILE")
stderr_bytes=$(file_bytes "$STDERR_FILE")
if [[ $((now - last_heartbeat_time)) -ge "$HEARTBEAT_SECONDS" ]]; then
heartbeat_count=$((heartbeat_count + 1))
append_status info in-progress "In progress ${heartbeat_count}"
last_heartbeat_time=$now
fi
if [[ "$stdout_bytes" -ne "$last_stdout_bytes" || "$stderr_bytes" -ne "$last_stderr_bytes" ]]; then
last_output_change_time=$now
stall_warning_logged=0
@@ -285,6 +327,7 @@ main() {
trap - EXIT
local final_stdout_bytes final_stderr_bytes
local success_file success_bytes
final_stdout_bytes=$(file_bytes "$STDOUT_FILE")
final_stderr_bytes=$(file_bytes "$STDERR_FILE")
@@ -294,6 +337,16 @@ main() {
exit 0
fi
if [[ ${#SUCCESS_FILES[@]} -gt 0 ]]; then
for success_file in "${SUCCESS_FILES[@]}"; do
success_bytes=$(file_bytes "$success_file")
if [[ "$success_bytes" -gt 0 ]]; then
append_status info completed "reviewer completed successfully via success file $(join_success_files)"
exit 0
fi
done
fi
append_status error completed-empty-output "reviewer exited successfully with empty stdout"
exit "$EXIT_COMPLETED_EMPTY_OUTPUT"
fi

View File

@@ -34,6 +34,9 @@ check_skill_file() {
assert_contains "$file" '$(cat /tmp/'
assert_contains "$file" "--strict-mcp-config"
assert_contains "$file" "--setting-sources user"
assert_contains "$file" "### P0"
assert_contains "$file" "In progress N"
assert_contains "$file" "notify-telegram.sh"
assert_not_contains "$file" "--allowedTools Read"
}

82
skills/reviewer-runtime/tests/smoke-test.sh Normal file → Executable file
View File

@@ -66,6 +66,28 @@ run_helper() {
return "$exit_code"
}
run_helper_allowing_success_file() {
local command_file=$1
local stdout_file=$2
local stderr_file=$3
local status_file=$4
local success_file=$5
shift 5
set +e
"$HELPER_PATH" \
--command-file "$command_file" \
--stdout-file "$stdout_file" \
--stderr-file "$stderr_file" \
--status-file "$status_file" \
--success-file "$success_file" \
"$@"
local exit_code=$?
set -e
return "$exit_code"
}
test_delayed_success() {
local dir=$1
local command_file=$dir/delayed-success.sh
@@ -120,6 +142,36 @@ printf "completed after soft timeout\n"
assert_file_contains "$status_file" "state=completed"
}
test_in_progress_heartbeats() {
local dir=$1
local command_file=$dir/in-progress-heartbeats.sh
local stdout_file=$dir/in-progress-heartbeats.stdout
local stderr_file=$dir/in-progress-heartbeats.stderr
local status_file=$dir/in-progress-heartbeats.status
make_command "$command_file" '
sleep 3
printf "finished with heartbeat coverage\n"
'
if run_helper "$command_file" "$stdout_file" "$stderr_file" "$status_file" \
--poll-seconds 1 \
--heartbeat-seconds 1 \
--soft-timeout-seconds 5 \
--stall-warning-seconds 4 \
--hard-timeout-seconds 10; then
local exit_code=0
else
local exit_code=$?
fi
assert_exit_code "$exit_code" 0
assert_file_contains "$stdout_file" "finished with heartbeat coverage"
assert_file_contains "$status_file" "state=in-progress"
assert_file_contains "$status_file" "In progress 1"
assert_file_contains "$status_file" "In progress 2"
}
test_nonzero_failure() {
local dir=$1
local command_file=$dir/nonzero-failure.sh
@@ -173,6 +225,34 @@ exit 0
assert_file_contains "$status_file" "state=completed-empty-output"
}
test_success_file_allows_empty_stdout() {
local dir=$1
local command_file=$dir/success-file.sh
local stdout_file=$dir/success-file.stdout
local stderr_file=$dir/success-file.stderr
local status_file=$dir/success-file.status
local success_file=$dir/review-output.md
make_command "$command_file" "
printf 'review body from redirected file\\n' > \"$success_file\"
exit 0
"
if run_helper_allowing_success_file "$command_file" "$stdout_file" "$stderr_file" "$status_file" "$success_file" \
--poll-seconds 1 \
--soft-timeout-seconds 5 \
--stall-warning-seconds 3 \
--hard-timeout-seconds 10; then
local exit_code=0
else
local exit_code=$?
fi
assert_exit_code "$exit_code" 0
assert_file_contains "$success_file" "review body from redirected file"
assert_file_contains "$status_file" "state=completed"
}
test_signal_cleanup() {
local dir=$1
local command_file=$dir/signal-child.sh
@@ -252,8 +332,10 @@ main() {
test_delayed_success "$tmp_dir"
test_soft_timeout_continues "$tmp_dir"
test_in_progress_heartbeats "$tmp_dir"
test_nonzero_failure "$tmp_dir"
test_empty_output_is_terminal "$tmp_dir"
test_success_file_allows_empty_stdout "$tmp_dir"
test_signal_cleanup "$tmp_dir"
test_hard_timeout_escalation "$tmp_dir"

View File

@@ -0,0 +1,158 @@
#!/usr/bin/env bash
set -euo pipefail
SCRIPT_DIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
HELPER_PATH=$(cd "$SCRIPT_DIR/.." && pwd)/notify-telegram.sh
fail() {
echo "FAIL: $*" >&2
exit 1
}
assert_equals() {
local actual=$1
local expected=$2
if [[ "$actual" != "$expected" ]]; then
fail "expected '$expected', got '$actual'"
fi
}
assert_file_contains() {
local file=$1
local pattern=$2
if ! grep -qF -- "$pattern" "$file"; then
echo "--- $file ---" >&2
sed -n '1,200p' "$file" >&2 || true
fail "expected '$pattern' in $file"
fi
}
capture_curl() {
local bin_dir=$1
local curl_args_file=$2
mkdir -p "$bin_dir"
cat >"$bin_dir/curl" <<EOF
#!/usr/bin/env bash
printf '%s\n' "\$@" >"$curl_args_file"
printf '{"ok":true}\n'
EOF
chmod +x "$bin_dir/curl"
}
read_curl_value() {
local file=$1
local key=$2
awk -v prefix="${key}=" 'index($0, prefix) == 1 { print substr($0, length(prefix) + 1); exit }' "$file"
}
test_missing_credentials() {
local tmp_dir=$1
local output_file=$tmp_dir/missing-credentials.out
set +e
env -u TELEGRAM_BOT_TOKEN -u TELEGRAM_CHAT_ID \
"$HELPER_PATH" --message "hello" >"$output_file" 2>&1
local exit_code=$?
set -e
[[ "$exit_code" -eq 2 ]] || fail "expected exit code 2 for missing credentials, got $exit_code"
assert_file_contains "$output_file" "bot token is required"
}
test_rejects_message_and_message_file_together() {
local tmp_dir=$1
local message_file=$tmp_dir/message.txt
local output_file=$tmp_dir/message-and-file.out
printf 'hello from file\n' >"$message_file"
set +e
TELEGRAM_BOT_TOKEN=test-token \
TELEGRAM_CHAT_ID=123456 \
"$HELPER_PATH" --message "hello" --message-file "$message_file" >"$output_file" 2>&1
local exit_code=$?
set -e
[[ "$exit_code" -eq 2 ]] || fail "expected exit code 2 for mutually exclusive arguments, got $exit_code"
assert_file_contains "$output_file" "use either --message or --message-file, not both"
}
test_successful_request() {
local tmp_dir=$1
local bin_dir=$tmp_dir/bin
local curl_args_file=$tmp_dir/curl-args.txt
capture_curl "$bin_dir" "$curl_args_file"
PATH="$bin_dir:$PATH" \
TELEGRAM_BOT_TOKEN=test-token \
TELEGRAM_CHAT_ID=123456 \
"$HELPER_PATH" --message "Plan completed"
assert_file_contains "$curl_args_file" "https://api.telegram.org/bottest-token/sendMessage"
assert_file_contains "$curl_args_file" "chat_id=123456"
assert_file_contains "$curl_args_file" "text=Plan completed"
assert_file_contains "$curl_args_file" "disable_web_page_preview=true"
assert_file_contains "$curl_args_file" "parse_mode=HTML"
}
test_message_file_and_custom_api_base() {
local tmp_dir=$1
local bin_dir=$tmp_dir/bin-message-file
local curl_args_file=$tmp_dir/curl-message-file.txt
local message_file=$tmp_dir/telegram-message.txt
capture_curl "$bin_dir" "$curl_args_file"
printf 'Plan completed from file\n' >"$message_file"
PATH="$bin_dir:$PATH" \
TELEGRAM_BOT_TOKEN=test-token \
TELEGRAM_CHAT_ID=654321 \
"$HELPER_PATH" \
--message-file "$message_file" \
--api-base-url "https://telegram.example.test/custom"
assert_file_contains "$curl_args_file" "https://telegram.example.test/custom/bottest-token/sendMessage"
assert_file_contains "$curl_args_file" "chat_id=654321"
assert_file_contains "$curl_args_file" "text=Plan completed from file"
}
test_truncates_long_message() {
local tmp_dir=$1
local bin_dir=$tmp_dir/bin-truncate
local curl_args_file=$tmp_dir/curl-truncate.txt
local long_message_file=$tmp_dir/long-message.txt
local truncated_message
capture_curl "$bin_dir" "$curl_args_file"
python3 - <<'PY' >"$long_message_file"
print("A" * 5000, end="")
PY
PATH="$bin_dir:$PATH" \
TELEGRAM_BOT_TOKEN=test-token \
TELEGRAM_CHAT_ID=123456 \
"$HELPER_PATH" --message-file "$long_message_file"
truncated_message=$(read_curl_value "$curl_args_file" "text")
assert_equals "${#truncated_message}" "4096"
}
main() {
[[ -x "$HELPER_PATH" ]] || fail "helper is not executable: $HELPER_PATH"
local tmp_dir
tmp_dir=$(mktemp -d)
trap "rm -rf '$tmp_dir'" EXIT
test_missing_credentials "$tmp_dir"
test_rejects_message_and_message_file_together "$tmp_dir"
test_successful_request "$tmp_dir"
test_message_file_and_custom_api_base "$tmp_dir"
test_truncates_long_message "$tmp_dir"
echo "PASS: telegram notifier tests"
}
main "$@"