The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple shell commands to manage the git lifecycle and interact with the GitHub API.
Uses git worktree and git branch for isolated development environments.
Uses the GitHub CLI (gh) to view repository metadata, monitor PR checks, fetch logs, and merge pull requests.
[EXTERNAL_DOWNLOADS]: The skill invokes bun install, which downloads project dependencies from the standard NPM registry when a bun.lock file is present.
[REMOTE_CODE_EXECUTION]: The skill runs local project scripts (bun test, bun run typecheck, bun run build) to validate changes. This executes code defined within the repository being processed.
[PROMPT_INJECTION]: The skill implements an automated feedback loop that parses external content from CI logs and PR review comments to generate code fixes, which is a surface for indirect prompt injection.
Ingestion points: Reads CI run logs via gh run view --log-failed and PR review bodies from cubic-dev-ai[bot] via gh api.
Boundary markers: No explicit boundary markers or delimiters are used to wrap the ingested log or review content before processing.
Capability inventory: The skill has the ability to modify the local filesystem, execute shell commands, and push code changes to remote repositories.
Sanitization: There is no evidence of sanitization, filtering, or validation performed on the ingested logs or review comments before they are used to influence the agent's implementation logic.

work-with-pr