菜单
Specification review skill for PAW workflow. Validates Spec.md against quality criteria and returns structured feedback for iteration.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/lossyrob/phased-agent-workflow --skill paw-spec-review命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Transforms evaluation findings into structured review comments with comprehensive rationale. Handles both initial draft generation and critique response iteration.
Posts finalized review comments to GitHub, leaving the review pending by default or submitting it under explicit verified authorization.
Analyzes PR changes to create ReviewContext.md and DerivedSpec.md artifacts. Handles both initial analysis and resumption after baseline research.
正在显示 SKILL.md
基于 SOC 职业分类
| name | paw-spec-review |
| description | Specification review skill for PAW workflow. Validates Spec.md against quality criteria and returns structured feedback for iteration. |
Execution Context: This skill runs in a subagent session, delegated by the PAW orchestrator. Return structured feedback (pass/fail + issues) to the orchestrator—do not make orchestration decisions.
Review specifications for quality, completeness, and clarity before planning proceeds. Return structured feedback—do not make orchestration decisions.
Reference: Follow Core Implementation Principles from
paw-workflowskill.
After review, the PAW agent receives:
Evaluate the specification at .paw/work/<work-id>/Spec.md against the Quality Criteria Checklist below. For failing items, note the specific issue, identify affected section(s), and suggest what needs to change without prescribing exact wording.
For passing specs: Indicate readiness for planning. Include minor polish suggestions if any.
For failing specs: List each failing criterion with the specific issue, affected section(s), and a suggestion for what to fix. Include count of passing vs total criteria.
Instead of: "FRs are incomplete" Say: "FR-003 lacks story mapping. Add (Stories: P2) reference."
Only flag issues that violate a checklist item. Personal preferences don't count unless they affect testability or clarity.
Identify what's wrong and suggest direction. Don't prescribe exact wording—the Spec Agent makes those decisions.
Consider workflow mode when reviewing:
Report to PAW agent: pass/fail result, criteria passing count, issues found count, and detailed feedback per Feedback Guidance above. Do NOT make orchestration decisions—return status only.
