Tools provided
Tools the agent exposes — your AI client calls them automatically when it needs them.
- assess_coverage — Work out which situations a change leaves unexercised — happy path, failure paths, input validation, state transitions — and rank the gaps by what each one would let through.read-only
- design_test_cases — Write the test cases a piece of behaviour needs — happy path, failures, validation, boundaries — each as setup, action and expected observable result, testing behaviour rather than implementation.read-only
- check_test_quality — Judge whether tests actually test anything — catching mirrored implementations, over-mocking, assertions that cannot fail, and tests that pass for the wrong reason.read-only
- plan_device_matrix — Plan the real-device conditions a mobile change must be exercised under — OS versions, permission states, network conditions, backgrounding, and the upgrade-from-previous-version case.read-only
- report_bug — Turn an observation into a bug report somebody can reproduce — steps, expected versus actual, scope, and an explicit separation of what was observed from what is suspected.read-only
- draft_test_report — Assemble the QA verdict — what was covered, what was found, what remains untested — reporting only results that were supplied and never implying a check that nobody ran.read-only
