flaky-test-analyzerlisted

# Playwright flaky test analyzer Find why a Playwright suite is unreliable, report it in priority order, fix what the user approves, and prove the fix with repeat runs. Target: $ARGUMENTS ## Working context - Playwright config in cwd: !`ls playwright.config.* 2>/dev/null || echo "none — search for it"` - Reporter output present: !`ls -d test-results playwright-report 2>/dev/null || echo "none"` ## Reference material Two files sit next to this skill. Read them before auditing — do not work from memory. - `references/rule-catalog.md` — the rules to check, with severity, detection method, signal, and rewrite for each. **This is the checklist you execute.** - `references/playwright-flakiness.md` — the explanation behind each rule ID, with worked before/after code. Read the relevant section before you rewrite anything, and cite the ID when you report a finding so the user can look it up. Rule IDs are stable (`A01`, `C04`, `D03`, …) and shared between both files. --- ## Step 1 — Scope Resolve the target from `$ARGUMENTS`: - A spec file → audit that file. - A directory → audit every `*.spec.*` / `*.test.*` under it. - `--report <path>` → see Step 2; the report determines the scope. - Nothing → read `playwright.config.*` and audit its `testDir`. Read `playwright.config.*` regardless of scope. It is itself an audit target (all of category F), and it tells you what the rest of the audit means: `fullyParallel`, `retries`, `trace`, `workers`, and the `use` block for E0