The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script collect_monorepo_evidence.py that utilizes subprocess.run to invoke git commands. The implementation uses argument lists rather than shell strings, which prevents command injection.
[COMMAND_EXECUTION]: The agent is instructed to run development commands such as make ebook and project-specific tests to maintain the monorepo's documentation and build integrity.
[DATA_EXFILTRATION]: The skill reads various project files, including AGENTS.md, package.json, and lock files to collect evidence. It includes a specific evidence policy to avoid the collection of secrets, production data, or internal content without a defined purpose.
[PROMPT_INJECTION]: The skill processes repository files like AGENTS.md as ingestion points for documentation. While explicit boundary markers are not present in the prompt, the agent's capabilities (executing git and make) are used within a controlled developer context on a verified repository, and the data collection script provides sanitization via structured JSON output.

specsfy-monorepo-documentator