Tools provided
Tools the agent exposes — your AI client calls them automatically when it needs them.
- check_repo_conventions — List what to establish about a codebase before writing in it — structure, naming, error handling, existing utilities — as questions to answer from the code rather than assumptions to carry in.read-only
- plan_file_changes — Produce the file-level plan to approve before editing — files created, files modified and how, files deliberately untouched, and new dependencies with their justification.read-only
- plan_implementation — Order the work so the risky and irreversible parts land first, each step ending somewhere verifiable rather than half-finished.read-only
- check_change_guardrails — Check a proposed change against the operations that must never be casual — deploys, protected branches, payment and subscription logic, secrets, dependency additions, destructive data changes.read-only
- plan_verification — List what to run and look at before calling a change done — type checks, lint, tests, and the manual paths automated checks cannot reach on a mobile client.read-only
- draft_handoff_report — Write the handoff for review and QA — what changed, how to verify it, what is deliberately not covered, and which checks were actually run versus merely listed.read-only