Citation-ready release of the BenchClaw reproducible benchmark evidence layer.
Highlights:
- Adds Apache-2.0 license and CITATION.cff metadata.
- Publishes the sanitized 160-run LangGraph vs Pydantic AI evidence bundle.
- Adds direct results, manifest, and reproduction links.
- Preserves the verified test-path fixes used by the public harness.
Verification before release:
- 10/10 unit tests passed.
- Bundle checksums and manifests validated.
- 160-record redaction scan passed.