post-feature-qa

name	post-feature-qa
description	Apply after EVERY feature implementation, story completion, or task finish. Act as a senior QA engineer: make sure all edge cases are covered and implementation fully aligns with the story/spec. Use whenever you complete a story, ship a feature, or before marking a task done or opening a PR. Do not skip this — run the QA checklist before claiming work is complete.

Purpose

Post-Feature QA — Senior QA Engineer Mode

When a feature, story, or task has been implemented, you MUST run this QA pass before marking it complete. Acting as a senior QA engineer, verify test coverage and story alignment. Never claim work is done without this review.

When to Trigger

• You just finished implementing a story or task
• You completed a feature and are about to commit or open a PR
• The user says "feature is done", "story complete", or "ready for review"
• You are about to mark checklist items complete in tasks/todo.md or similar
• Before any "verification complete" or "acceptance criteria verified" claim

Core Workflow

1. Load the Story/Spec

• Find the story or feature spec. Common locations:
  • ruleforge-prd.md (or similar PRD)
  • tasks/todo.md
  • docs/, specs/, or ticket/issue description
• Extract:
  • Acceptance criteria — what must be true for the work to be done
  • Task list — individual work items (e.g. 2.1.1, 2.1.2)
  • Explicit test requirements — e.g. "Write tests: valid, invalid, every field type"

2. Cross-Check Acceptance Criteria

For each acceptance criterion:

Criterion	Action
Behavioral	Verify a test exists that asserts the behavior
Config/format	Verify tests for valid + invalid inputs
Error handling	Verify tests for error paths and error messages
Edge cases	Empty, null, boundary values, invalid types

If any criterion has no corresponding test, add the test.

3. Edge Case Checklist

Load .agentsmesh/skills/post-feature-qa/references/edge-case-checklist.md for a full reference. For every public API, config schema, or behavior:

• Valid inputs: minimal valid, full valid, optional fields omitted
• Invalid types: wrong type (string vs number, array vs object, etc.)
• Boundaries: empty array/string, single element, max length if applicable
• Null/undefined: missing required fields, explicit null where relevant
• Invalid values: out-of-range, invalid enum, malformed structure
• Combined failures: multiple invalid fields in one input

Apply this systematically to each field or function you implemented.

4. Story Alignment

• Every task in the story has corresponding code or tests
• No scope creep — nothing implemented that wasn't in the spec
• Exported types/APIs match what the spec describes
• Defaults, error messages, and formats match the spec

5. Run Verification

• Run the test suite: pnpm test (or equivalent)
• Integration tests: Run pnpm build && pnpm test — integration tests in tests/integration/ require the built dist
• Run lint: pnpm lint
• Optionally: pnpm test --coverage and check coverage for new code
• Fix any failures before claiming complete

6. Integration Test Checklist

When completing a story, ensure integration tests exist and pass:

• Story 1–2: tests/integration/story-1-cli.test.ts, story-2-config.test.ts
• For new stories: Add story-N-<name>.test.ts covering acceptance criteria end-to-end
• Integration tests run the full CLI (spawn) or full config/generation pipeline with fixtures

Output Format

Produce a QA Report with:

## QA Report — [Story/Task ID]

### Acceptance Criteria

| Criterion | Covered by test? | Status   |
| --------- | ---------------- | -------- |
| ...       | ✓ / ✗            | OK / GAP |

### Edge Cases

| Scenario | Covered? | Test location |
| -------- | -------- | ------------- |
| ...      | ✓ / ✗    | path:line     |

### Gaps Identified

- [ ] Missing test for X
- [ ] No test for invalid Y

### Actions Taken

- Added tests for ...
- Fixed ...

If gaps exist, fix them and re-run the report until all gaps are closed.

Principles

• Evidence over assertions: Do not say "tests cover edge cases" without showing which tests cover which cases.
• Explicit over implicit: List each edge case and the test that covers it.
• Fix, don't report: When you find a gap, add the test. Don't just note it and move on.
• Story as source of truth: If the spec says "every field type", ensure every field has invalid-type tests.