Recall Regression Evidence
This deterministic corpus checks whether every configured scanner rule still detects a known-positive fixture and ignores a paired negative fixture. It is a regression suite, not an empirical estimate of real-world recall or safety.
12scanner rules
24synthetic fixtures
24passing fixtures
Not measuredempirical recall
Rule coverage
Interpretation boundary
- This internally-authored synthetic corpus is a deterministic scanner regression suite, not an independent empirical recall measurement.
- Passing fixtures does not estimate the probability that a real Skill is safe or that novel malicious behavior will be detected.
- Empirical recall remains not measured until a release-bound corpus is labelled independently of scanner output and kept blinded during evaluation.