The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes git log --oneline -20 to understand recent project changes for documentation updates. This is a common and safe practice for development tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the codebase to generate documentation, which represents an injection surface.
Ingestion points: Reads README.md, configuration files (package.json, Cargo.toml), and source directories (SKILL.md).
Boundary markers: The instructions include a safeguard to "Focus on facts" and "Do not invent requirements," and provide an Agent Context Guide that instructs the agent to use a question tool if documentation conflicts with intent.
Capability inventory: Performs file system reads, file system writes within the docs/ directory, and limited git command execution.
Sanitization: No specific string sanitization is mentioned, but the workflow emphasizes factual extraction rather than open-ended execution.

docs