Description
Agent skill repositories are installed by thousands of people who never open them. I linted 1,676 skill files from popular repositories this month: 333 errors and 589 warnings, mostly broken references, unknown frontmatter fields and argument hints that reference nothing. Skills like that silently do less than their README promises.
This service checks a repository of Claude Code or agent skills (SKILL.md, CLAUDE.md, AGENTS.md, hooks, MCP configs) with a linter and then by hand, and returns what is actually broken and how to fix it.
What you receive within 3 business days
- The full lint report (agnix) with every finding, grouped by rule and file.
- A human review of the twenty most consequential findings: why each one matters at runtime and the exact fix.
- A size and structure check of your memory files against the recommended token limits, because a 17,000-token CLAUDE.md is a rulebook nobody follows.
- A security pass on hooks and MCP configuration for tool poisoning patterns.
Limits
One public or shared repository, up to 300 markdown files. I do not rewrite your skills for you; I tell you what to change and where. No execution of your code: static review only.
🚀 Level Up Your AI Game
Get weekly AI tools, prompts & automation strategies — free, every week.
No spam. Unsubscribe anytime.
