Execute an implementation plan end-to-end: reads a plan file, implements each task with build/test cycles, runs acceptance validation, and reports readiness. Use when you have a written plan and want to execute it with quality gates.
Purpose: Execute a written implementation plan end-to-end with per-task build/test cycles and a final acceptance gate.
<path> — path to the plan file (optional; if omitted, finds most recent file in docs/plans/)Find the plan file:
.md file in docs/plans/ (by filename date prefix or modification time).Parse the plan into an ordered task list. Tasks are typically numbered sections, headings, or checklist items in the plan.
Report: "Found N tasks in <plan-file>. Starting execution."
For each task in order:
Read the task description from the plan. Understand what files to create/modify and what behavior is expected.
Follow the plan exactly. Create/modify files as specified. Do not add unplanned features or refactors.
Run the project build command (from CLAUDE.md or package.json):
pnpm -r build
Run tests for changed packages:
pnpm --filter <changed-package> test -- --run
Apply these code-review specialists against the files changed in this task:
code-review/specialists/typescript-qualitycode-review/specialists/securityFor each finding:
Max 3 auto-fix cycles per finding set.
Create a commit with a message referencing the plan task:
<type>(<scope>): <description>
Task N from <plan-file>
Before moving to the next task, output:
Task N/M: DONE — <brief summary>
Files: <list>
Commit: <sha>
If a task fails after retries, output:
Task N/M: BLOCKED — <failure description>
And stop execution.
After all tasks are complete:
Apply ALL code-review specialists against the cumulative changeset:
git diff origin/main --name-only
For each finding:
git diff origin/main).Invoke the /checkpoint skill (lint, build, test) as the final quality gate.
Output a structured report:
## Execution Report
**Plan:** <plan-file>
**Tasks:** N/N completed
### Per-Task Summary
| # | Description | Commit | Files Changed | Status |
|---|-------------|--------|---------------|--------|
| 1 | ... | abc1234 | 3 | DONE |
| 2 | ... | def5678 | 2 | DONE |
### Code Review Findings
- Blocker: N (N auto-fixed)
- High: N (N auto-fixed)
- Medium: N
- Low: N
### Plan Alignment
- All tasks implemented: YES/NO
- Unplanned changes: NONE / [list]
- Spec deviations: NONE / [list]
### Checkpoint
- Lint: PASS/FAIL
- Build: PASS/FAIL
- Tests: PASS/FAIL
### Verdict: PASS / WARN / FAIL
[If PASS]: Ready to ship. Run /ship or /pr to continue.
[If WARN]: Review Medium/Low findings before shipping.
[If FAIL]: Fix listed issues before shipping.