Summary

• Document the generic WP Codebox benchmark contract and caller-owned scoring boundary.
• Add bench summarize and artifacts bench-results helpers for extracting benchmark summaries from recipe-run output or artifact bundles.
• Cover the helper output with a fixture-only smoke test.

Closes #572.

Verification

• npm run build
• npm run benchmark-summary-smoke
• npm run command-registry-smoke
• npm run discovery-command-smoke

AI assistance

• AI assistance: Yes
• Tool(s): OpenCode (GPT-5.5)
• Used for: Implemented benchmark helper CLI plumbing, documentation draft, and smoke coverage; Chris remains responsible for review and merge.