The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard command-line tools like gh and git to inspect repository metadata (visibility) and state. It also recommends running local project formatters like prettier to validate documentation files. These tools are used appropriately within the scope of security analysis and documentation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of repository source code and configuration to identify threats. It mitigates potential injection risks by providing specific subagent briefs that require strict evidence citations (files and functions) and by using structured templates for the final threat model document.
[SAFE]: The skill incorporates several high-standard security practices:
It explicitly warns against writing secrets, credentials, or internal hostnames into the documentation.
It requires checking the repository's public visibility before recording open threats to prevent accidental disclosure.
It instructs the agent to treat data from neighboring systems as untrusted input rather than verified truth.
It limits its own scope to reading code and configuration, specifically stating it does not send requests to deployed environments or commit changes automatically.

threat-model