Agent Data Sandbox
Internal validation prototype for the agent & agency reference-data catalogue (TA-001 + Addendum A, Draft). Search and inspect synthetic agents, agencies, credentials, source evidence and review states.
Cases where automation stopped. Reviewer outcomes below update local demonstration state only — no external service is called and no real decision is recorded.
No source is production-approved. States below mirror the governed Source Registry (Feasibility Notes v1.1): priority does not equal approval, WNBPA remains blocked pending an authorised source, the UK register is unidentified, and RealGM is excluded absent permission or a licence.
All twelve governed edge cases, repeatable and safe for automated testing. Counts: . Every reference carries a SYN- marker.
Pilot allocation (ceilings): FIBA 35 · NBPA 30 · BBL/DBB 20 · FIP 15 = 100; UK may replace part once identified and approved; WNBPA/RealGM never substituted. Access-control check (28 Jul 2026): this staging host answers HTTP 200 with no Cloudflare Access challenge — it is public, so real-person records must not ship in this downloadable bundle. Loading additionally requires per-source compliance clearance and Legal approval of the limited validation-processing basis (register #12). Collection plan: docs/technical-architecture/pilot-100-record-collection-plan.md. Founder-supplied known-scenario records appear only as fictionalised analogues (tagged on the records).
Replays Dataset A through the governed pipeline in this browser: staging → provenance → normalisation → resolution → review routing → publication gate. Counts are computed live from the fixtures — deterministic, offline, no external calls.