What problem does it solve? When a Playwright E2E suite rots, manually triaging dozens of failing tests is slow and error-prone. This Skill autonomously baselines the suite, classifies failures as deterministic or flaky, and dispatches one repair worker per failing spec file until every test is green or carries a documented explanation. ## Core Features & Use Cases - Baseline and classification: Runs a discovery pass plus focused failure reruns to distinguish deterministic failures from flaky tests, respecting @known-defect tags. - Parallel per-file repair workers: Fans out one worker per red file following a gated pipeline (reproduce, evidence-analysis, context-probe, experiment, understand, fix, verify) with a one-way understand gate before any fix. - Audit-grade reporting: Writes report.md and schema-validated report.json under .achilles/self-repair/<run-id>/ with per-test outcomes, bug evidence bundles, and a complete NDJSON event log. - Use Case: After a nightly CI run turns red, run npm run test:repair to have each failing spec diagnosed and healed in parallel, with app bugs reported alongside slow-motion reproduction recordings instead of silently patched tests. ## Quick Start Ask the agent to self-repair the suite, or run npm run test:repair to launch the autonomous per-file repair pipeline.
