The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard tools including the GitHub CLI (gh) and Git to manage pull requests and repository state. It includes a custom shell script (scripts/stop-worktree-server.sh) designed to identify and terminate local development processes (via portless) that are tied to specific Git worktrees. The script includes safety checks to ensure only processes within the target worktree are affected.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external data by reading pull request descriptions and repository metadata. Although this creates a surface for indirect prompt injection, the risk is mitigated by instructions to verify remote states and a general lack of dangerous autonomous capabilities that could be exploited via text injection.
[SAFE]: The skill incorporates security-conscious instructions, specifically requiring the agent to inspect and redact credentials, personal data, and private information from screenshots and videos before they are uploaded as GitHub attachments.

merge