The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a bundled Python script (audit-docs.py) that uses the subprocess module to run local git commands for verifying repository history. These commands are executed without a shell and use paths validated against the repository root, effectively mitigating command injection and directory traversal risks.
[INDIRECT_PROMPT_INJECTION]: The documentation audit functionality reads and processes content from Markdown files within the repository, creating a potential surface for indirect injection.
Ingestion points: The audit scripts (audit-docs.py and module_audit.py) recursively read all .md files in the repository.
Boundary markers: No explicit delimiters are used when presenting the results of the audit to the agent.
Capability inventory: The agent can execute the audit script, which has read access to the local filesystem and git history, but lacks network and file-write permissions.
Sanitization: The script extracts specific metadata using regular expressions rather than processing the entire text as natural language. However, any malicious text contained within identified links or identifiers could be included in the report, potentially influencing the agent's context.

docs-governance