菜单
Audits and optimizes a collection of agent skills for compliance, concision, and cross-tool portability. Applies Agent Skills spec, Anthropic best practices, and progressive disclosure principles. Use when reorganizing skills, auditing skill quality, or optimizing a skill catalog.
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/Harduex/skills --skill optimizing-skill-repo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。
The universal base layer of standards for everything a project authors, code and prose alike, derived from the ASD-STE100 controlled-language specification: a controlled naming vocabulary with a per-project glossary, small imperative units with guard conditions first, severity-accurate errors that state consequences, and a document band covering procedural and descriptive writing, sentence-length counting, punctuation, and cohesion. A project's own coding-standards skill and glossary complement and override it. Use when writing or naming code, authoring comments, docstrings, error messages, logs, or commit messages, writing or rewriting a README, specification, runbook, release note, or any technical document, auditing code or a document against a standard, rewriting someone else's prose into the standard, setting up or extending a project glossary, or when the user mentions coding standards, naming conventions, writing style, documentation style, or controlled vocabulary.
Converts an internal artifact — a design spec, a shipped branch or feature, an API or data contract — into a self-contained handoff document an external team can read fast and act on without any internal context. Use when handing a contract, integration guide, or feature description to another team or company, when asked for a handoff document, a client-facing explainer, or a doc to publish outside the repo (Google Doc, wiki, PDF), when converting a spec into something reviewable by people without codebase access, or when folding a published copy's review edits back into the repo source.
Distributes uncommitted working-tree changes back to their originating commits on a multi-commit branch using git commit --fixup and a non-interactive autosquash rebase. Use when the user wants to "distribute" fixes into existing commits, clean up a branch after review feedback, fold review fixups into the right commits, or mentions fixup/autosquash workflows.
基于 SOC 职业分类
正在显示 SKILL.md
| name | optimizing-skill-repo |
| description | Audits and optimizes a collection of agent skills for compliance, concision, and cross-tool portability. Applies Agent Skills spec, Anthropic best practices, and progressive disclosure principles. Use when reorganizing skills, auditing skill quality, or optimizing a skill catalog. |
Systematically audit and improve a skill collection. Interview the user at each decision point — provide recommended answer, wait for confirmation before proceeding.
Optimization checklist:
- [ ] Phase 1: Structure — flatten, remove scaffolding, decide repo purpose
- [ ] Phase 2: Inventory — classify each skill (keep/merge/drop/rewrite)
- [ ] Phase 3: Research — fetch latest spec + best practices
- [ ] Phase 4: Audit — check every skill against principles
- [ ] Phase 5: Rewrite — apply fixes, deduplicate, add examples/checklists
- [ ] Phase 6: Verify — final compliance pass
Resolve these decisions with the user:
For each skill, classify:
Present classification table. User approves before proceeding.
Fetch latest standards:
Extract actionable principles (see Audit Principles below).
Check every skill against:
Spec compliance:
name matches directory, lowercase + hyphens, ≤64 charsdescription ≤1024 chars, third person, "Use when [triggers]"Content quality:
Effectiveness patterns:
Cross-skill health:
Present audit table with issues per skill.
Apply fixes in batch:
Final pass: