菜单
Use after executing a task directory - verifies implementation against plan.md, acceptance checks, and anti-goals
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| 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.md from disk.spawn_agent (dispatch in parallel when needed).<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>
<the_process>
Read the active task directory completely.
For each acceptance check in plan.md:
Search for:
If there are gaps, lead with findings. If approved, cite the evidence that supports approval. </the_process>
Use when shaping a feature before implementation - turns rough ideas into an approved markdown spec and local task directory
Use to execute an approved local markdown task directory iteratively - work the current Now slice, update docs, then stop
Use when implementation is complete and verified - delete finished local task docs, present integration options, and clean up safely
Use for advanced markdown task-doc operations - split work, reprioritize slices, delete finished local task dirs, and keep docs clean
Use when starting any conversation - establishes skill selection and the local task-doc workflow
Use before claiming work complete, fixed, or passing - requires fresh evidence before any completion claim