The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use find, grep, and wc for mapping codebase structure and searching for code patterns within the repository.
[PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by processing content from all repository source code files.
Ingestion points: All project source files (via find, glob, and grep).
Boundary markers: None identified.
Capability inventory: Writing analysis reports to the .audits/ directory on the local filesystem.
Sanitization: None identified.
[SAFE]: The skill is a legitimate analysis tool that operates locally within the project directory. No evidence of network exfiltration, credential theft, or unauthorized persistence was found. The identified injection surface is low-risk given the restricted capabilities.

architectural-analysis