feat(do-task): add claude-code variant SKILL.md + template (M1+M2)

M1 canonical specs (opencode reviewer research, task-plan template, Review Loop subroutine, Phase 1-10 prose, secret-scan regex list) are embedded in the M2 claude-code SKILL.md as the canonical reference; later variants (M3 codex, M4 cursor, M5 opencode) will fork from this file. Reviewer: codex / gpt-5.4. Approved after 3 rounds: - Round 1: 2 P1 + 3 P2 → REVISE - Round 2: 1 P2 → REVISE - Round 3: 0 P0/P1/P2, 1 P3 (non-blocking) → APPROVED Key design properties: - Plan-review payload strips Runtime State and Review History to prevent reviewer session-ID leakage across rounds. - Secret-scan step 1a redacts matched text to [REDACTED:<pattern-label>:<match-length>-chars] before any user surfacing; never echoes raw match content. - Brainstorming required for any behavior-changing task; auto-skip limited to pure-documentation and pure-comment-whitespace-rename. - Phase 3 reviewer config defaults to codex / gpt-5.4 with MAX_ROUNDS=10 when user opts for defaults; explicit interactive default also gpt-5.4 for internal consistency. - Template Metadata records Branch Name and Worktree Path. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-22 14:13:25 -05:00
parent c344e96984
commit 437b2024cd
2 changed files with 892 additions and 0 deletions
@@ -0,0 +1,144 @@
+# Task Plan: [Short Title]
+
+> **Variant guardrail (Claude Code):** When generating or updating this file, the agent MUST be out of plan mode. Sub-skills (`brainstorming`, `test-driven-development`, `verification-before-completion`, `finishing-a-development-branch`, `using-git-worktrees`) MUST be invoked through the `Skill` tool explicitly — no shell wrappers.
+
+## Metadata
+
+| Field | Value |
+|-------|-------|
+| Created | YYYY-MM-DD |
+| Slug | YYYY-MM-DD-<slug> |
+| Runtime | claude-code |
+| Reviewer CLI | codex \| claude \| cursor \| opencode |
+| Reviewer Model | <model> |
+| MAX_ROUNDS | 10 |
+| Branch Strategy | current-branch \| worktree |
+| Branch Name | <current branch name, or new branch name when worktree is used> |
+| Worktree Path | <absolute path to worktree dir; blank when Branch Strategy = current-branch> |
+| Status | draft |
+
+### Status Enum (authoritative)
+
+| Value | Meaning |
+|-------|---------|
+| `draft` | Newly created; plan review not yet started |
+| `plan-approved` | Plan review loop returned APPROVED |
+| `implementation-in-progress` | Phase 6 executing |
+| `implementation-approved` | Phase 8 review loop returned APPROVED; awaiting commit |
+| `pushed` | Committed + pushed to remote |
+| `local-only` | Committed locally; user declined push |
+| `aborted-plan-review` | MAX_ROUNDS reached in Phase 5; user aborted |
+| `aborted-impl-review` | MAX_ROUNDS reached in Phase 8; user aborted |
+| `aborted-verification` | Phase 7 retries exhausted; user aborted |
+| `failed` | Hard tooling failure |
+
+---
+
+## Prompt
+
+<!-- Exact user prompt, verbatim. -->
+
+## Interpretation
+
+<!-- Short restatement of goal + out-of-scope items. -->
+
+## Assumptions
+
+<!-- Anything we're assuming and needs confirmation. Empty list OK after clarifying questions. -->
+
+## Files
+
+<!-- Files expected to be created / modified / deleted. Paths are absolute or repo-relative. -->
+
+| Action | Path | Why |
+|--------|------|-----|
+|        |      |     |
+
+## Approach
+
+<!-- 3-10 bullets describing implementation order. -->
+
+## TDD Approach
+
+<!-- One of:
+  (a) **TDD applies** — list the failing test(s) to write first, then implementation, then confirm green.
+  (b) **TDD auto-skipped** — reason must be exactly one of:
+      - `pure-documentation`
+      - `pure-comment-whitespace-rename`
+  (c) **TDD user-approved skip** — user explicitly approved skipping TDD for this task.
+      Record the approval timestamp (ISO-8601) and the specific reason (e.g., `pure-config-addition`).
+-->
+
+## Acceptance Criteria
+
+- [ ] <criterion 1>
+- [ ] <criterion 2>
+
+## Verification
+
+<!-- Commands to run:
+  lint: <cmd>
+  typecheck: <cmd>
+  tests: <cmd>
+-->
+
+## Rollback
+
+<!-- How to undo: `git revert <hash>`, or manual steps if the change is not easily revertable. -->
+
+---
+
+## Runtime State
+
+<!-- Updated by the skill at runtime. Used to detect resume and to persist reviewer session IDs across rounds. -->
+
+```yaml
+plan_review_round: 0
+implementation_review_round: 0
+CODEX_PLAN_SESSION_ID:
+CODEX_IMPL_SESSION_ID:
+CURSOR_PLAN_SESSION_ID:
+CURSOR_IMPL_SESSION_ID:
+OPENCODE_PLAN_SESSION_ID:
+OPENCODE_IMPL_SESSION_ID:
+last_phase_entered:
+last_round_ts:
+last_scan_outcome_plan:
+last_scan_outcome_impl:
+verification_attempts: 0
+tests_added_count: 0
+tdd_used: false
+```
+
+## Review History
+
+<!-- Append one entry per reviewer round, both loops. -->
+
+| Timestamp (ISO-8601) | Loop | Round | Verdict | Summary |
+|----------------------|------|-------|---------|---------|
+|                      |      |       |         |         |
+
+## Final Status
+
+<!-- Filled at the terminal outcome (phase 9/10). Populate at least:
+  - Terminal status (one of the 10 Status enum values)
+  - Commit hash (if any)
+  - Plan-review rounds used / MAX_ROUNDS
+  - Implementation-review rounds used / MAX_ROUNDS
+  - TDD used (true|false)
+  - Tests added count
+  - Verification attempts used
+  - Last round ISO-8601 timestamp
+  - Notes (anything the user should know when revisiting)
+-->
+
+
+---
+
+## Guardrails (do NOT remove)
+
+- This file is the single persistent artifact for `do-task`. Do not split it or delete it on success.
+- `Status` must always match one of the 10 enum values.
+- `Runtime State` is updated by the skill, not by the user.
+- Review History is append-only.
+- `last_scan_outcome_plan` and `last_scan_outcome_impl` record the most recent secret-scan result for each loop. They are informational; the scan itself runs per-payload with no caching.