d9c16fb914
Block C implementation: - adversarial_cases.yaml: 30 tricky cases in 5 categories (wrong legal basis, dark patterns, incomplete docs, similar-but-different, homonyms) - test_adversarial.py: 63 tests validating adversarial cases - test_regression.py: ontology stability, dependency engine, quality metrics - conftest.py: shared fixtures (DB session, sample controls) Total: 371 tests passing (221 existing + 150 new). Real-world benchmarks (C1) need manual ground truth creation. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>