The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes system commands to integrate with git for version control and file management.
Evidence: The helper script scripts/check_docs.py uses subprocess.run to execute git rev-parse and git ls-files commands.
Context: These operations are performed safely by passing arguments as a list rather than a single string, preventing shell injection vulnerabilities. The commands are limited to legitimate repository metadata and file listing tasks.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, user-controlled markdown files which could potentially contain malicious instructions.
Ingestion points: Documentation files are read by the check_docs.py script and an auditing subagent to identify duplication, broken links, and structural issues.
Boundary markers: The skill design enforces a strict human-in-the-loop gate, requiring explicit user approval of a proposed plan before any file modifications (Write, Edit, or Git operations) are executed.
Capability inventory: The skill possesses the ability to modify the file system via Write and Bash (specifically for git mv and git rm) based on its analysis of the ingested data.
Sanitization: The skill does not explicitly sanitize file content but relies on its structural analysis purpose and human verification to mitigate the risk of the agent following instructions embedded within the documentation.

tidy