What problem does it solve? Test suites that drift slow or flaky erode trust in the gate until developers stop running it. This Skill owns the repository's test suites: it keeps the local gate under a 120-second budget, tracks which suites are parked and why, and replaces guesswork about slowness with measured per-class timings. ## Core Features & Use Cases - Fast local gate enforcement: Runs scripts/test-local.ps1 as the authoritative gate, kills suites exceeding the 120-second budget, and treats overruns as a parking decision rather than a test failure. - Parked suite management: Tracks which suites are excluded from the default run, the measured cost of each, and the batch-by-batch process (at most 100 individually verified tests) for bringing them back. - Bad test elimination: Maintains a named list of intermittently failing tests and requires each to be rewritten deterministically or deleted, with retries and quarantines explicitly forbidden. - Measurement-driven diagnosis: Provides TRX-based per-class timing recipes and records confirmed costs and false trails so prior investigations are not repeated. - Use Case: When a developer reports "tests are slow" or the gate goes red intermittently, invoke this Skill to read the handover state, identify the bad tests or slow classes by measurement, and fix or park them without raising the budget. ## Quick Start Ask the agent to run the test-manager skill to diagnose why the test gate is slow and continue the queued test cleanup work.
