# Agent system card

## Purpose and scope

- User outcome and conventional baseline:
- In-scope and out-of-scope tasks:
- Users, tenants, regions, and environments:

## Architecture and authority

- Data, control, effect, and observability planes:
- Models, prompts, tools, stores, protocols, and runtimes:
- Per-action autonomy and permission table:
- Human checkpoints and irreversible effects:

## Evidence

- Evaluation contract/report:
- Threat model and privacy assessment:
- SLO and capacity model:
- Safety case and risk-register entry:

## Operations

- Owners and escalation path:
- Version bundle and deployment topology:
- Monitoring, rollback, kill switch, and recovery:

## Residual risk

- Known limitations and unresolved hazards:
- Risk owner, acceptance rationale, and expiration: