The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions and compatibility section recommend executing a local script, scripts/repo_readme_probe.py, to gather repository evidence. The accompanying test script, scripts/test_repo_readme_probe.py, uses subprocess.run to call this probe script using the local Python interpreter (sys.executable). These operations are local, scoped to the repository being analyzed, and do not involve shell execution of unsanitized user input.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external repository files, such as existing READMEs, package manifests, and CI configurations. This presents an inherent surface for indirect prompt injection if those files contain malicious instructions. However, the skill provides specific instructions to the agent to distinguish between repository documentation and agent operating rules, emphasizing that factual manifests should take precedence over prose.
[DATA_EXPOSURE]: The repo_readme_probe.py script reads project manifests and configuration examples (e.g., .env.example). It includes protective measures such as directory pruning for sensitive or large folders like .git, .venv, and node_modules to ensure it only inventories structural metadata rather than sensitive data.

repository-readme-writer