Summary

• add evaluation release-binding coverage to model-supply-chain so model artifacts are tied to immutable evaluation dataset revisions/checksums, thresholds, run IDs, evaluator identity, environment, and release decisions
• distinguish model cards from release-quality evaluation evidence
• expand backdoor detection with canary, trigger, and targeted-slice regression evidence bound to artifact ID, dataset version, run ID, and timestamp
• add Not Applicable / Not Evaluable handling for private evaluation data or model classes where trigger tests do not apply
• update severity criteria, report template, maturity summary, common pitfalls, tags, and changelog for version 1.0.1

Scope

This addresses #1171. I also posted an attempt comment before implementation: #1171 (comment)

Closes #1171

/claim #1171

Validation

• git diff --check (only existing Windows LF-to-CRLF warning)
• verified markdown code fence count is even (16)
• verified issue-specific markers for evaluation integrity, evaluation release binding, immutable dataset evidence, release-result binding, canary/trigger regression, artifact-to-evaluation binding, Not Evaluable handling, and version 1.0.1