The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs project reconnaissance using ls, find, and cat to discover the tech stack and existing settings. It also recommends providing the agent with broad access to the gh api tool.
[EXTERNAL_DOWNLOADS]: The skill suggests including the @linear/mcp-server package in the project's Model Context Protocol configuration.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes external dependency and configuration files which are managed by the project owners and could contain instructions targeting the auditor.
Ingestion points: package.json, pyproject.toml, Gemfile, Cargo.toml, and other project metadata files.
Boundary markers: None identified.
Capability inventory: File discovery and read operations (ls, find, cat).
Sanitization: None identified.

claude-settings-audit