The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill utilizes the --admin flag with the gh pr merge command to perform merges. This capability allows the agent to bypass repository branch protection rules, such as required reviews or specific status checks, effectively escalating the agent's authority to bypass standard security gates.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits a significant attack surface for indirect prompt injection by processing untrusted data.
Ingestion points: The skill reads repository-level configuration files (AGENTS.md, CLAUDE.md) and fetches external review comments from sources like Greptile, CodeRabbit, and human reviewers via the GitHub API (gh pr view --comments).
Boundary markers: There are no explicit boundary markers or instructions to ignore embedded commands within the ingested content.
Capability inventory: The skill possesses extensive capabilities including modifying files, pushing code to remote branches (git push), and merging PRs (gh pr merge).
Sanitization: No sanitization or validation logic is defined to prevent instructions within comments or metadata files from influencing the agent's behavior.
[COMMAND_EXECUTION]: The skill performs autonomous command execution through tools like git, gh (GitHub CLI), and uv run lintro. It is instructed to intelligently resolve merge conflicts and fix CI failures by executing generated code and tests within the worktree.
[EXTERNAL_DOWNLOADS]: The skill interacts with external network resources via the GitHub CLI and API to fetch logs (gh run view), triage comments, and verify check statuses from remote CI providers.

babysit-pr