The Agent Skills Directory

[COMMAND_EXECUTION]: The script fetch.py invokes the gh (GitHub CLI) tool to perform GraphQL queries. It uses subprocess.run with arguments passed as a list, which is a secure practice that prevents shell injection vulnerabilities. The execution is limited to the system's installed GitHub client for its primary function.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content from GitHub issues and pull requests to calculate triage priority.
Ingestion points: Untrusted data enters the skill via GitHub API responses in fetch.py (specifically issue and PR titles, bodies, and comments fetched in paginate_nodes).
Boundary markers: The script does not use explicit boundary markers or delimiters when handling the text content of issues, as it is primarily used for ranking logic rather than re-interpolation into further prompts within the script itself.
Capability inventory: The skill has the capability to execute the gh binary via subprocess.run for data retrieval across multiple scripts (e.g., fetching comments, commits, and reviews).
Sanitization: The script uses regular expressions (like STAMP_RE and CLOSING_LIST_RE) to strictly filter and extract specific metadata (triage stamps and closing keywords) from the untrusted text rather than treating the text as executable instructions.

Triage eligible fetch