1. Run Drift Detection

Purpose: Run all drift detection tests — taxonomy, schema, structure, voice, and difficulty drift — plus orchestrator integration.

Prerequisites: None (regex-based, no API keys or DB needed)

Cost / Duration: $0 | ~10-30 seconds

Prompt

Run the complete drift detection test suite:

```bash
bun test packages/ai/src/drift/
```

This covers 6 drift test files:
- `taxonomy-drift.test.ts` — Taxonomy vocabulary and prefer-over violations
- `schema-drift.test.ts` — Schema conformance drift
- `structure-drift.test.ts` — Structural format drift
- `voice-drift.test.ts` — Voice and tone drift
- `difficulty-drift.test.ts` — Difficulty calibration drift
- `orchestrator.test.ts` — Drift orchestrator integration

Plus taxonomy-specific suites:
- `taxonomy-isolation.test.ts` — No cross-contamination between taxonomies
- `taxonomy-integration.test.ts` — End-to-end taxonomy pipeline

Plus FCG group validation:
- `group-diversity-validation.test.ts` — FCG key diversity (min unique keys, max per key, difficulty sequence)

All tests must pass. Any failure indicates prompt or rule drift that
needs investigation before further content generation.

Verification

  • All drift test files pass (exit code 0)
  • No skipped tests
  • Taxonomy isolation confirms zero cross-contamination
  • Schema drift confirms all fact keys match current fact_record_schemas
  • Voice drift confirms tone markers are consistent
  • Group diversity validation passes (FCG key diversity checks)

Back to index