For agent platforms and developers
See what your AI agent can do - and what it actually does.
An agent's real authority is spread across runtime configuration, tools, plug-ins, sandboxes, approval rules and deployment settings. Standard Bearer turns them into one Authority Map.
Four ideas in 60 seconds
- OpenClaw Observer
- OpenClaw is the first adapter, not the product boundary. Today the observer replays recorded OpenClaw configurations offline; it does not yet connect to a running agent or enforce policy.
- Effective authority
- What the agent can actually exercise: filesystem, shell, elevated execution, network and messaging, across every control that applies.
- Authority Exposure Gap
- Where effective authority exceeds what was intended. When the observer cannot see part of the picture, the gap is reported as unknown, never as none.
- Drift Planned
- What changed, and when. Drift calculation is defined in the architecture but not yet implemented.
Run the local scenarios
Both run offline from the repository. Installation and usage are documented in the README; they are not repeated here.
Scenario 06: Sandbox / elevated execution mismatch
Reveals hidden effective authority in an OpenClaw agent: a deployment expected to stay in /workspace that can also reach the shell, elevated execution and the open internet.
standard-bearer replay 06-sandbox-mismatchExample 03: Financial authority Simulation only
A customer delegates one utility payment with a $500 cap per billing cycle. Deterministic and offline: no bank, no payment rail, no money movement, self-attested evidence.
standard-bearer replay financial-authority