The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the GitHub CLI (gh) to retrieve issue data and post comments. These commands are executed as part of the intended workflow and are restricted to standard issue management tasks.
[PROMPT_INJECTION]: The skill handles untrusted data from GitHub issues, which presents a surface for indirect prompt injection. * Ingestion points: Issue descriptions and user comments are fetched in Step 1 via gh issue view or WebFetch. * Boundary markers: There are no specific delimiters or instructions to ignore instructions embedded within the issue body. * Capability inventory: The agent is capable of running gh commands to read/write data and is directed to trace logic through the codebase. * Sanitization: No automated sanitization is specified; however, the risk is mitigated by mandatory human-in-the-loop (HITL) confirmation at Steps 1, 3, 4, and 5 before any actions or classifications are finalized.

issue-triage