| Secure runtime | Staging environment, HTTPS intake/return, authenticated requests, and network boundaries. | Authorized endpoint test, request correlation, invalid-request rejection, and replay checks. | Planned |
| Identity & secrets | Least-privilege access, tenant boundaries, managed credentials, and rotation. | Role, revocation, cross-tenant denial, and key-lifecycle checks in the deployed environment. | Planned |
| Durable evidence | Persistent decision records, evidence storage, agreed retention, and integrity-preserving exports. | Restart persistence, authorized retrieval, tamper detection, and a reviewer-readable export. | Planned |
| Observation | Authorized evidence collection with explicit source, version, timestamp, and independence boundaries. | A traceable artifact from the approved workflow, plus visible handling of missing evidence. | Pilot gate |
| Operations | Structured logging, monitoring, alert routing, incident escalation, and support ownership. | Trace one request across the system and exercise an agreed failure/escalation scenario. | Planned |
| Recovery | Backups, known-good baselines, restore checks, and accountable recovery review. | A restore drill tied to the correct baseline, evidence revision, and human disposition. | Pilot gate |
| Production release | Separate staging/production, hardened integrations, change control, and security review. | Independent assessment, release/rollback evidence, and acceptance of the exact operating scope. | After pilot |