The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from external sources (CI logs and GitHub reviews) and uses this information to drive an automated 'fix and re-push' loop, creating a surface for potential instruction injection.
Ingestion points: The skill fetches CI failure logs using gh run view --log-failed and retrieves bot comments via gh api "repos/${REPO}/pulls/${PR_NUMBER}/reviews". This content is directly used to inform Phase 1 implementation fixes.
Boundary markers: No specific delimiters or boundary instructions are defined to separate external log/review content from the agent's internal reasoning or instructions.
Capability inventory: The skill possesses significant capabilities including pushing code to repositories (git push), merging PRs (gh pr merge), installing packages (bun install), and managing the file system (git worktree, mkdir, cp, rm).
Sanitization: While the instructions explicitly require sanitizing logs before writing them to a PR body (Phase 2), there is no evidence of sanitization or validation of the incoming data from CI or Cubic reviews before the agent processes it to generate new code.
[COMMAND_EXECUTION]: The skill relies extensively on shell command execution to manage the development lifecycle.
Evidence: Uses git, gh CLI, and bun to perform repository management, environment setup, and code verification.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of dependencies from well-known registries.
Evidence: Executes bun install during worktree setup to prepare the environment for implementation and testing.

work-with-pr