| name | review-implementation |
| description | Use after executing a task directory - verifies implementation against plan.md, acceptance checks, and anti-goals |
<codex_compat>
This skill was ported from Claude Code. In Codex:
- "Skill tool" means read the skill's
SKILL.md from disk.
- "TodoWrite" means create and maintain a checklist section in your response.
- "Task()" means
spawn_agent (dispatch in parallel when needed).
- Claude-specific hooks and slash commands are not available; skip those steps.
</codex_compat>
<skill_overview>
Review the implementation against the approved spec, not just against what happened to get coded.
</skill_overview>
<rigidity_level>
LOW FREEDOM - Review with evidence. Do not approve based on intuition or diff-skimming.
</rigidity_level>
<quick_reference>
| Step | Action |
|---|
| 1 | Read plan.md, context.md, and tasks.md |
| 2 | Verify acceptance checks with evidence |
| 3 | Search for anti-goal violations and unfinished work |
| 4 | Report findings or approve with evidence |
Evidence: file references, command output, or explicit test results.
</quick_reference>
<when_to_use>
- Before claiming feature completion
- Before deleting a finished local task directory
- After a meaningful execution slice if risk is high
</when_to_use>
<the_process>
1. Load the docs
Read the active task directory completely.
2. Check the acceptance contract
For each acceptance check in plan.md:
- find direct implementation evidence
- run the relevant verification command if needed
- mark gaps explicitly
3. Check anti-goals
Search for:
- forbidden shortcuts
- TODO or placeholder remnants
- behavior that contradicts the spec
4. Report clearly
If there are gaps, lead with findings.
If approved, cite the evidence that supports approval.
</the_process>
