The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs shell command execution to automate development workflows.
Uses git for branching, worktree management (git worktree add/remove), and atomic commits.
Uses gh (GitHub CLI) for PR creation, status checks (gh pr checks), and merging.
Uses bun for running tests, type checks, and build processes.
[EXTERNAL_DOWNLOADS]: The skill downloads external dependencies and interacts with remote services.
Executes bun install to fetch Node.js packages if a lockfile is detected.
Communicates with GitHub via Git and the GitHub CLI to sync code and manage PR resources.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface through the ingestion of external data.
Ingestion points: Reads TASK_SUMMARY for branch naming and retrieves bot review comments from the GitHub API via gh api in SKILL.md.
Boundary markers: None present to isolate interpreted instructions from data.
Capability inventory: High-privilege command execution capabilities including file system modification, network access via CLI tools, and test execution.
Sanitization: Input for branch names is sanitized using tr to ensure shell-safe characters; review logic uses username filtering (cubic-dev-ai[bot]) to verify the source of approval signals, which effectively mitigates common third-party spoofing attempts.

work-with-pr