The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a bash script template (templates/docgate.sh) designed for repository auditing. It utilizes standard Unix utilities including awk, grep, sed, and git. Command injection risks are mitigated by restricting input to git commands (e.g., commit SHAs) to specific regex patterns limited to hex characters.
[INDIRECT_PROMPT_INJECTION]: The auditing script processes markdown files that could potentially contain untrusted data. Ingestion points: Markdown files within the docs/ directory and the repository root. Boundary markers: The strip_asides function uses awk to remove fenced code blocks and HTML comments from the text being analyzed. Capability inventory: Execution of shell commands via git, awk, and grep within the template script. Sanitization: Input is normalized (whitespace adjustment, emphasis removal) and filtered by awk logic before regex patterns are applied. The deterministic nature of the script's checks significantly limits exploitation risk.
[SAFE]: All external URLs and resources reference official vendor infrastructure (ssheleg) or well-known services. The skill promotes robust documentation security practices, including append-only logs and verification of claims against immutable version control history.

evidence-docs