菜单
Use when creating new skills, editing existing skills, auditing skill quality, converting skills to markdown conventions, or verifying skills before deployment. Triggers include skill authoring requests, skill review needs, or "the skill doesn't work" complaints.
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
Lightweight implementation orchestrator for low-complexity work — fixes, refactors, doc changes, or single-AC features that do not warrant a phase plan or factory decomposition.
Linear phase-loop orchestrator for single-feature implementation plans. Use for medium-complexity work where transparent human-in-the-loop phase review is preferred over factory automation.
Implementation entry point. Use to execute a completed specification. Auto-detects the decomposition tier (Direct, Incremental, or Factory) from spec artifacts and dispatches to the matching execution sub-skill.
Decompose a single-feature specification into a linear, phase-by-phase implementation plan. Use this for medium-complexity work — single feature, one or two components — where transparent human-in-the-loop phase review is preferred over factory automation.
Scaffold, status-check, and manage specification directories. Use when creating a new spec, reading spec status, transitioning between phases, or logging decisions on a spec in .start/specs/.
Create a comprehensive specification from a brief description. Runs requirements gathering, solution design, and decomposition — routing decomposition to one of three tiers based on a complexity classifier: Direct (no plan), Incremental (linear phase plan), or Factory (parallel units with holdout scenarios).
| name | writing-skills |
| description | Use when creating new skills, editing existing skills, auditing skill quality, converting skills to markdown conventions, or verifying skills before deployment. Triggers include skill authoring requests, skill review needs, or "the skill doesn't work" complaints. |
Act as a skill authoring specialist that creates, audits, converts, and maintains Claude Code skills following the conventions in reference/conventions.md.
Request: $ARGUMENTS
SkillAuditResult { check: string status: PASS | WARN | FAIL recommendation?: string }
State { request = $ARGUMENTS mode: Create | Audit | Convert skillPath: string type: Technique | Pattern | Reference | Coordination }
Always:
Never:
If you catch yourself thinking any of these, STOP and follow the full workflow:
| Rationalization | Reality |
|---|---|
| "I'll just create a quick skill" | Search for duplicates first |
| "Mine is different enough" | If >50% overlap, update existing skill |
| "It's just a small change" | Small changes break skills too |
| "I can see the fix is correct" | Test it anyway |
| "The pattern analysis shows..." | Analysis != verification |
| "No time to test" | Untested skills waste more time when they fail |
match ($ARGUMENTS) { create | write | new skill => Create mode audit | review | fix | "doesn't work" => Audit mode convert | transform | refactor to markdown => Convert mode }
Search existing skills:
plugins/*/skills/*/SKILL.mdVerify frontmatter: Read first 10 lines — valid YAML? name + description present?
Verify structure: Grep for ## headings — PICS sections present?
Verify size: Line count < 500? If not, identify content to externalize.
Verify conventions: Read reference/conventions.md and check compliance.
For discipline-enforcing skills: Launch Task subagent with pressure scenario per reference/testing-with-subagents.md.
Format report per reference/output-format.md.
match (mode) { Create => steps 2, 3, 7 Audit => steps 4, 7 Convert => steps 5, 7 }