Test plan

Plan type test. You run in the MAIN session and interview the human. The testing contract — coverage matrix, pyramid, evidence rules, the playwright-cli track and the declarative E2E scenarios — is references/testing.md; read it before planning.

Procedure

1. Language — resolve lang (references/language.md).
2. Read the code under test first, plus context/testing.md (frameworks, real commands, tests_dir, E2E dirs) and context/conventions.md. No testing context → suggest feat-map.
3. Choose the plan kind with the human:
   • Create tests (IMPLEMENT — tests are code and are committed): §4 lists the test files, the E2E scenario files and the specs exported from them; §5 maps each behavior/AC to UNIT-/INT-/E2E- IDs.
   • Audit coverage (REPORT): §4 lists only .planning/feat/features/{slug}/test-audit.md, rendered from templates/test-report.md (not report.md: some runtimes refuse subagent writes to report-like names).
4. Method — follow references/pwdevia-method.md with plan_type: test.
5. Focus
   • Persona: QA engineer / test specialist on the project's frameworks.
   • Priority: business logic → edge cases (null, empty, boundary) → error paths → security (authn/authz, input sanitization) → critical UI flows.
   • Unit is never NOT_APPLICABLE; integration, E2E, accessibility and responsiveness only with a written justification.
   • UI flows: one scenario file per flow with the variants happy path, validation error, forbidden, empty state; Verification includes python3 "<plugin-root>/scripts/e2e_scenarios.py" validate <scenarios-dir> --plan <plan.md> and the export --to playwright command into <tests-dir>/e2e/generated.
   • Interactive UI checks use playwright-cli in the session feat-{slug}, run from the feature's evidence/ directory, with one screenshot per UI criterion.

Prohibitions

• Never skip E2E scenarios when the code under test has UI.
• Never test framework internals or write assertions that only check "is defined".
• Never put literal credentials in scenarios — ${env:NAME} placeholders only.

Language: resolve lang per references/language.md before any human-facing output. Safety: never read or expose .env* (except .env.example/.template/.sample), keys, certificates or credentials — references/safety.md. Paths references/, scripts/, templates/, schemas/ are relative to the plugin root (references/runtime.md).