The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python's subprocess module to interact with Git and the GitHub CLI (gh) to manage repository forks and submissions.
Evidence: scripts/submit_contribution.py and scripts/star_repo.py execute commands such as git commit, git push, and gh pr create.
Mitigations: Commands are executed using argument lists rather than shell strings, preventing shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill clones the project's repository to a temporary directory to facilitate the creation of pull requests.
Evidence: scripts/submit_contribution.py runs gh repo clone Aperivue/medsci-skills.
Source: The download targets the official repository of the skill's author on GitHub, which is a well-known and trusted service.
[DATA_EXFILTRATION]: The skill transmits local modifications to GitHub to fulfill its primary purpose of contributing to the project.
Evidence: scripts/submit_contribution.py pushes code changes to the user's fork and opens pull requests.
Mitigations: The skill implements a multi-layered safety check. It uses scripts/check_contribution_safety.py to scan for Protected Health Information (PHI) and hardcoded secrets, blocking submission if blockers are found. It also mandates a manual review where the user must inspect every line of the diff before confirming the submission.
[INDIRECT_PROMPT_INJECTION]: The skill processes local files that may contain untrusted data if modified by other processes or agents.
Ingestion points: Local skill files read by scripts/find_local_changes.py.
Boundary markers: The skill instructions emphasize human review as the final safety gate.
Capability inventory: Network access via gh/git and local file writes to the qc/ directory.
Sanitization: A specialized safety scanner validates all content before it is prepared for submission.
[CREDENTIALS_UNSAFE]: A hardcoded dummy API token is present in the test suite for validation purposes.
Evidence: tests/test_contribution_safety.sh contains the string ghp_abcdefghijklmnopqrstuvwxyz0123456789.
Context: This is a placeholder value used exclusively for regression testing of the secret detection logic in the safety scanner.

contribute