e2e-runner

Quick Summary

Goal: Generate and maintain E2E tests on the project's auto-detected framework with mandatory TC-code↔test traceability, producing a suite that runs, repeats deterministically, and provably maps every test back to its spec.

Summary:

• Read the project E2E reference + config FIRST — detect the framework (Playwright/Cypress/Selenium) before writing anything.
• Every test carries a TC-{MODULE}-E2E-{NNN} code — traceability back to the spec is mandatory.
• Use Page Object Model; derive stable selectors from data-testid/BEM/ARIA, never positional/generated ones; keep assertions out of page objects.
• Deterministic runs — unique GUID test data, explicit waits only, reused auth session; after fixing failures, record the learning back in the E2E reference doc.

Workflow:

1. Read project E2E docs — docs/project-reference/e2e-test-reference.md, docs/project-config.json
2. Detect framework — Playwright, Cypress, Selenium, etc.
3. Load test specs — find TC codes in feature docs
4. Generate/update tests — follow Page Object pattern
5. Run tests — project's configured commands
6. Update docs — add learnings to docs/project-reference/e2e-test-reference.md

Key Rules:

• NEVER fabricate file paths, function names, or behavior — investigate first, then act — why: a hallucinated selector or path passes review and fails only at runtime
• MUST ATTENTION include TC code (TC-{MODULE}-E2E-{NNN}) in every test — why: traceability is how a test proves which spec it covers
• NEVER hardcode brittle selectors — derive stable ones from data-testid, BEM classes, or ARIA roles — why: generated/positional selectors break on every unrelated markup change
• ALWAYS update visual baselines when UI changes — why: stale baselines turn real regressions into false failures
• MUST ATTENTION read the project E2E reference BEFORE any E2E work — why: local conventions override generic framework defaults

---

Evidence Gate — Every claim, finding, and recommendation requires file:line proof or traced evidence. Confidence >80% to act; <80% must verify first. NEVER speculate without proof. External Memory — For complex or lengthy work (research, analysis, scan, review), write intermediate findings and final results to a report file in plans/reports/ — prevents context loss.