The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted third-party content from GitHub pull request metadata and diffs. The agent context includes instructions to treat this data as untrusted and disregard any embedded commands. The ingestion points are gh pr list and gh pr diff in SKILL.md. Boundary markers are present in the safety model, requiring the agent to ignore instructions in external data. Capability inventory includes repository write operations such as merging and branch deletion. The risk is mitigated by mandatory user review of a consolidated merge plan and explicit confirmation gates before any action is taken.
[COMMAND_EXECUTION]: The skill uses git, gh, and jq for repository management. These commands are standard for the tool's purpose. The execution is protected by confirmation gates for all destructive or significant operations, ensuring that the agent does not perform unintended repository modifications.

merge-open-prs