The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to interact with git and the GitHub CLI (gh) for cloning repositories, committing changes, and managing pull requests. These calls use argument lists, which is a safe practice that prevents shell injection.- [DATA_EXFILTRATION]: The skill is designed to send local modifications to a public GitHub repository. Security is maintained through a mandatory PII/PHI safety scan (check_contribution_safety.py) and a hard requirement for the user to review and confirm the complete diff before transmission.- [CREDENTIALS_UNSAFE]: The file tests/test_contribution_safety.sh contains a hardcoded dummy string ghp_abcdefghijklmnopqrstuvwxyz0123456789. This value is used solely as a test case to verify that the safety scanner correctly identifies and blocks GitHub personal access tokens.- [PROMPT_INJECTION]: The skill processes untrusted local data (user-edited files) that could contain indirect prompt injections. Evidence Chain: (1) Ingestion points: Reads modified skill files from the agent's local directory. (2) Boundary markers: Explicit instructions to display the full text for user confirmation. (3) Capability inventory: Network access via gh and git for PR creation. (4) Sanitization: Mandatory safety scan for PII and secrets, though it does not specifically filter for instructional override patterns.

contribute