Evidence, with the coverage boundary left on.
Release checks prove specific code paths and executor scorecards summarize independently verified receipts. Neither is a promise of future execution, a security audit, or authority for an agent to act.
Release-candidate evaluation
Recorded 29 July 2026 for the release candidate that produced this page. The rendered commit above binds the deployment to its source revision. Commands are reproducible from the repository; opt-in fork checks are reported separately so a hermetic pass is never misrepresented as live-chain coverage.
872 Vitest checks passed; lint, TypeScript, and the production Next.js build completed.
148 Node checks passed across execution, receipts, relay intake, private submission, adapter provenance, notifications, SDK typed-data safety, MCP protocol, and package safety.
422 Foundry checks passed with five explicitly opt-in skips; the hermetic CI partition passed 335 with four skips.
87 fixed-block Base and Robinhood fork checks passed, plus four explicitly enabled Robinhood end-to-end gates.
The PostgreSQL 16 migration/concurrency harness, production dependency audit, and full-history secret scan passed.
Receipt-backed executor scorecard
Look up an executor address. The result counts only finalized or reverted executions nominated to the reference receipt pipeline and re-verified against canonical factory-created Zaps. It is not a global reputation oracle, identity proof, or permission grant.
How to interpret this
- A passing release check establishes only the named command and scope at the recorded revision.
- Reliability is finalized attempts divided by all recorded attempts, expressed in basis points.
- Missing receipts mean missing coverage, not evidence that an executor did or did not act.
- Scorecard reads are public and have
authorityScope: none.