name: qa-spec-auditor description: Living documentation auditor that compares requirements vs UI/behavior vs API contract vs test results to detect spec drift, outdated docs, and undocumented features. output_dir: reports/audit dependencies: recommended: - qa-requirements-generator - qa-spec-writer

QA Spec Auditor

Purpose

Compare "how it should be" vs "how it is" after each release or sprint. Detect spec drift, outdated documentation, undocumented features, and regression set optimization opportunities. Produce actionable drift reports with task suggestions for qa-task-creator.

Trigger Phrases

• "Spec audit" / "Documentation audit for [release/sprint]"
• "Spec drift" / "Detect drift between docs and implementation"
• "Outdated documentation" / "Docs vs implementation comparison"
• "Undocumented features" / "Find features not in specs"
• "Requirements vs UI" / "Requirements vs API contract"
• "Living documentation audit" / "Spec compliance check"
• "Regression set optimization" / "Worn out tests" / "Test rotation"

Comparison Dimensions

See references/audit-checklist.md for checklists per dimension.

Detection Categories

See references/drift-patterns.md for common patterns and remediation.

Output Deliverables

1. Drift Report — Per dimension: findings, severity, affected artifacts
2. Action Items — Update docs / fix implementation / create tests
3. Task Suggestions — Structured for qa-task-creator (doc updates, test creation, spec fixes)
4. Regression Set Recommendations — Tests to rotate, prioritize, or deprioritize

Drift Report Template

# Spec Audit — [Release/Sprint]

## Summary
| Dimension | Findings | Critical | Action Items |
|-----------|----------|----------|--------------|
| Req ↔ UI | N | X | [Count] |
| Req ↔ API | N | X | [Count] |
| Req ↔ Tests | N | X | [Count] |
| Docs ↔ Impl | N | X | [Count] |

## Findings by Dimension
### Requirements ↔ UI
- [Finding] — [Severity] — [Action]
### Requirements ↔ API
- [Finding] — [Severity] — [Action]
### Requirements ↔ Tests
- [Finding] — [Severity] — [Action]
### Documentation ↔ Implementation
- [Finding] — [Severity] — [Action]

## Task Suggestions (for qa-task-creator)
1. [Task type] — [Description] — [Owner]
2. ...

## Regression Set Recommendations
- Rotate: [Tests that are "worn out"]
- Prioritize: [High-value tests]
- Deprioritize: [Low-value, stable tests]

Workflow

1. Input: Requirements, specs, OpenAPI/API docs, test cases/RTM, optional UI URL
2. Requirements ↔ UI: Use Playwright MCP to explore UI; compare flows, elements, validation to spec
3. Requirements ↔ API: Compare OpenAPI/spec to requirements; detect undocumented endpoints, schema drift
4. Requirements ↔ Tests: Validate RTM; find untested requirements, orphan tests
5. Documentation ↔ Implementation: Compare docs to code/UI/API; flag outdated sections
6. Regression optimization: Analyze test history; identify always-pass, low-value tests; suggest rotation
7. Output: Drift report, action items, task suggestions for qa-task-creator

MCP Tools Used

• Playwright MCP (cursor-ide-browser): UI exploration, compare live behavior to spec
• GitHub MCP: Change tracking, PR/commit history for docs and code

Integration with Other Skills

Scope

Can do (autonomous):

• Compare requirements vs UI (via Playwright MCP)
• Compare requirements vs API contract (OpenAPI)
• Compare requirements vs test results (RTM, coverage)
• Compare documentation vs implementation
• Detect outdated docs, undocumented features, spec drift
• Produce drift report with action items
• Suggest tasks for qa-task-creator
• Recommend regression set optimization (test rotation)
• Use GitHub MCP for change tracking

Cannot do (requires confirmation):

• Modify documentation or code directly
• Change requirements without stakeholder approval
• Override organizational doc/spec policies

Will not do (out of scope):

• Execute tests (consume results only)
• Write or modify test automation
• Deploy or change production systems

Quality Checklist

• All four comparison dimensions addressed
• Drift report includes severity per finding
• Action items are specific (update X, fix Y, create test Z)
• Task suggestions formatted for qa-task-creator
• Regression recommendations include rationale
• No hardcoded secrets; URLs/config referenced

Troubleshooting

Reference Files