Florian BRUNIAUX
|
a55ff38143
|
feat(quiz): add complete audit system for 256 questions
Implements automated pipeline for quiz question quality control:
**Phase 1: Context Extraction**
- Script: extract-audit-context.py
- Resolves doc_reference anchors to guide sections (97.3% success)
- Multi-file support (ultimate-guide.md, learning-with-ai.md, etc.)
- Fuzzy matching + substring fallback
- Output: audit-context.json (256 questions + context)
**Phase 2: Batch Generation**
- Script: generate-audit-batches.py
- 16 prioritized review batches by category
- Advanced Patterns split into 2 batches (29 questions)
- Embedded review instructions in each batch
- Output: audit-batches/*.md (16,559 lines)
**Phase 3: Report Compilation**
- Script: generate-audit-report.py
- Parses agent review outputs (PASS/ISSUE format)
- Aggregates by severity (critical/warning/info)
- Output: audit-report.md
**Validation:**
- Q01-001 error found immediately (curl vs npm contradiction)
- System working as designed ✅
**Documentation:**
- AUDIT-WORKFLOW.md (complete 5-phase guide)
- AUDIT-SYSTEM-SUMMARY.md (architecture + metrics)
- IMPLEMENTATION-COMPLETE.md (status + validation)
- DEMO-REVIEW-OUTPUT.txt (example review)
**Next Steps:** Manual agent reviews (16 batches, ~2-3 hours)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
|
2026-02-04 16:45:47 +01:00 |
|