Playbooks
How to implement — paired with Blueprints for reference designs.
Eval engineering
The eval engineering playbook series (golden datasets, plane playbooks, judge calibration, and more) is in draft — visible in local dev, not yet published to production.
Start with G.A.I.N Evaluation for principles; the full blueprint and playbooks link from Eval Blueprint.
PGAR runtime
Four playbook groups for Policy-Governed Agent Runtime. Start at the PGAR Runtime overview.
| Group | Start here | Covers |
|---|---|---|
| Foundation | Overview | SARAC, token custody, PEP/PDP, step-up, audit |
| Assurance | Policy test scenarios | CI authorization cases, adversarial bypass |
| Boundary | Overview | Five trust boundaries in request order |
| Domain | Tool registry | Tools, manifests, RAG retrieval |
Also see PGAR Blueprint for the reference design. Bridge reading: PGAR with RAG. Eval overlap: Action plane and Tool plane.