| Skill | Description | Install |
|---|---|---|
| agent-linter | Audits all the context an AI agent receives (AGENTS.md, CLAUDE.md, skills, referenced docs, MCP configs) and generates an executive report of contradictions, ambiguities, redundancies, and structural issues. | npx skills add https://github.com/mathiew82/skills --skill agent-linter |
npx skills add https://github.com/mathiew82/skills --skill <skill-name>
Example:
npx skills add https://github.com/mathiew82/skills --skill agent-linter
npx skills add https://github.com/mathiew82/skills --all
Once installed, invoke the skill from your agent (Claude Code, Cursor, etc.) by its name.
Example:
run agent-linter on this project
The skill will analyze your agent's context and generate a report. It will not modify any file unless you explicitly ask it to.
agent-linter generates an executive report by default. You can ask for other formats in plain language — no need to memorize commands.
- Detailed report: ask for "detail" or "a detailed report". You'll get the full version with the original passages side by side, so you can decide which rule to keep.
Example:
run agent-linter and give me the detailed report - Plain language report: ask for "plain language", "no jargon", or "explain like I'm not a developer". You'll get the same executive report but without technical terms.
Example:
run agent-linter but explain it in plain language - Apply changes: after the report, you can reply with "apply all", "apply only [section]", or "apply one by one". The skill will never modify anything unless you ask it to.
You don't need to know any specific commands. Just ask for what you want in your own words. The skill detects the intent, not the exact phrasing.
skills/
├── README.md
├── LICENSE
├── .gitignore
└── skills/
└── agent-linter/
├── SKILL.md
└── prompts/
├── analyze-pair.md
├── cluster-rules.md
└── build-report.md
Each skill lives in its own folder under skills/. A skill is defined by its SKILL.md file, which contains the frontmatter (name, description, license) and the instructions the agent follows.
This is a personal collection. If you find a bug or have a suggestion, feel free to open an issue.
MIT — see LICENSE for details.