automagik-hive: Command for Claude Code

/wish-review – Wish Completion Audit

---

description: Execute an end-to-end review loop for a wish, combining test execution, Death Testament aggregation, and scoring.

[CONTEXT]

• Designed for branches where multiple forge tasks have landed and the wish is nearing closure.
• Runs a standardized audit so humans can iterate until the wish earns a 100/100 readiness score.
• Operates outside the sandbox: command may trigger UV tests, diff inspection, and report generation.

[SUCCESS CRITERIA] ✅ Parses the wish file for status, success criteria, and linked artefacts. ✅ Aggregates latest commits plus Death Testament references into the audit report. ✅ Executes each --tests command (default set provided below) via uv run … and captures pass/fail. ✅ Emits a structured review summary, including completion score (0–100), blockers, and next-step recommendations. ✅ Records evidence paths (wish, forge plan, DTs, test logs) for traceability.

[DEFAULT TEST MATRIX]

• uv run pytest tests/lib/auth/test_credential_service.py -q
• uv run pytest tests/integration/auth/test_single_credential_integration.py -q
• uv run pytest tests/cli/test_docker_manager.py -q

[COMMAND SIGNATURE]

/wish-review @path/to/wish.md \
    [--tests "<command>"]... \
    [--score-only]

• Required first argument: wish file, referenced with @ so the report can link back.
• --tests may be repeated to override or extend the default matrix.
• --score-only skips command execution and recomputes score using cached results (for fast follow-up after minor doc edits).

[OUTPUT]

• Human-readable report (markdown) summarizing:
  • Wish metadata (status, phases, success criteria)
  • Commits since last review (with task IDs)
  • Test results table (pass/fail + command output snippet)
  • Completion score (0–100) with rubric: implementation, validation, outstanding risks
  • Recommended follow-up forge tasks (auto-generated when score < 100)
• Optional JSON payload for scripts integrating with future automation.

[PROCESS BREAKDOWN]

1. Discovery – Load wish, forge plan, and latest Death Testaments; compute change summary.
2. Validation – Run queued test commands (unless --score-only), capture logs.
3. Scoring – Apply rubric; highlight blockers and residual risks.
4. Recommendation – Suggest new forge task groups or close-out actions based on score.
5. Report – Emit markdown + JSON (under tmp/wish-review-<timestamp>.json).

[NOTES]

• Always use uv run … for Python tooling.
• Completion score of 100/100 requires: all success criteria met, tests passing, no unresolved risks.
• When blockers exist, include evidence and reference any new Forge tasks created in response.