audit-assistance-provenance

name	audit-assistance-provenance
description	Audit a run, strategy, tool, dataset, prompt, or report for hidden operational assistance H, knowledge assistance K, engine leakage, retries, and attribution errors.

Audit assistance and provenance

Audit path

1. Trace every input visible before the final action.
2. Classify parser/rules/environment/legal-actions/model-only-search/engine involvement as H0-H7.
3. Classify generic/phase/opening/retrieved/current-position knowledge as K0-K7.
4. Compare declared ceiling to effective events.
5. Inspect SAN +/#, top-k, evaluations, PVs, candidate labels and tool outputs for leakage.
6. Inspect SDK/runtime for hidden retry, fallback, routing or cache.
7. Verify dataset/training-derived engine information is distinguished from live assistance.
8. Verify report and leaderboard names match the system, not only the base model.
9. Emit violations with evidence event/artifact IDs.
10. Block claim/publication when attribution cannot be reconstructed.

Output

assistance_audit:
  declared: {R: R3, H: H3, K: K4}
  effective: {R: R3, H: H5, K: K4}
  status: protocol_violation
  findings: []