菜单
Use after producing a deliverable and before handoff to run a fixed-rubric self-critique and revise cycle. Reduces hallucination and drift.
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
Use for general product implementation work that is not primarily backend architecture, pure integration wiring, or screenshot-driven design-to-code.
Use when the main deliverable is maintainable documentation such as repository rules, onboarding guides, runbooks, ADRs, or architecture notes.
Use on first entry to a new repository to run environment scanning and ask targeted boundary questions before implementation.
Use after writing or modifying code to enforce the mandatory write → test → fix → repeat validation cycle.
Use before committing to a design or plan to force assumption-surfacing. The agent challenges your design, questions edge cases, and flags gaps — you patch vague decisions. Prevents the failure mode where a design "feels explained" but contains hidden flaws that only appear during implementation.
Use to establish and maintain a shared domain glossary (UBIQUITOUS_LANGUAGE.md). Creates a single source of term definitions that all agents, prompts, and documents must use — preventing semantic drift and repeated re-explanation across sessions.
| name | self-reflection |
| description | Use after producing a deliverable and before handoff to run a fixed-rubric self-critique and revise cycle. Reduces hallucination and drift. |
Use this skill to validate your own output before emitting a deliverable or handing off to another role. This is an intra-role behavior — each role critiques its own work using a fixed rubric, then revises before finalizing.
This is distinct from the critic role, which is an inter-role mechanism where one agent challenges another agent's plan.
Run self-reflection after producing a draft deliverable and before emitting the final output or handoff artifact. This applies to every role that produces structured output (plans, implementations, reviews, documentation).
| Scale | Required rubric dimensions | Max revision rounds |
|---|---|---|
| Small | Correctness + Adherence (2 of 5) | 1 |
| Medium | All 5 dimensions | 1 |
| Large | All 5 dimensions | 2 |
Evaluate your draft output against these five dimensions. For each, state pass or fail + specific issue.
DECISIONS.md?semi-auto) or an unintended violation?After evaluating, produce a brief reflection block (do not omit even if all pass):
## Self-reflection
- Correctness: [pass | fail — description]
- Consistency: [pass | fail — description]
- Adherence: [pass | fail — description]
- Completeness: [pass | fail — description]
- Isolation: [pass | fail | skipped — description]
- Revisions made: [list of changes, or "None"]
Do not enter an infinite revision loop. The maximum rounds are defined by task scale (see table above).
When this skill produces a reflection result, include it in the handoff artifact under state.reflection_result (if using the structured handoff schema) or as an appendix to the text-based handoff artifact.
Before marking self-reflection as complete, verify: