The Agent Skills Directory

[COMMAND_EXECUTION]: The bundled audit script scripts/check_supply_chain.py executes the GitHub CLI (gh) via subprocess.run to retrieve and verify repository security settings and vulnerability alerts.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to download and run security utilities like pip-audit and cyclonedx-py using uvx or pip, while explicitly recommending the use of pinned versions to mitigate the risk of executing untrusted code.
[INDIRECT_PROMPT_INJECTION]: The audit script scripts/check_supply_chain.py identifies a potential attack surface by reading untrusted repository configuration files; however, the risk is limited as the script is read-only and employs conservative regex-based inspection.
Ingestion points: The script reads .github/dependabot.yml, lockfiles (uv.lock, poetry.lock), .github/workflows/*.yml, CODEOWNERS, and pyproject.toml.
Boundary markers: No explicit delimiters are used to separate untrusted file content from internal processing logic.
Capability inventory: The script has the capability to execute shell commands via the gh CLI.
Sanitization: The script uses errors='replace' during file reads and employs regular expressions for parsing rather than executing or fully interpreting the data.

python-supply-chain