CLAW-ROOM.OS
Private browser scorecard / one workflow

Find the failure path before your agent finds it.

Answer six questions about one live or launch-ready workflow. You will get a fast control read, not a vague maturity score. No answers leave this browser.

Six-point control scan

Choose the answer that describes the workflow today.

3 min / 6 checks
1. Does the agent touch customer messages, money, or production data?

Risk rises when an action can create an external or irreversible consequence.

2. Is a human approval gate required before risky sends, spends, or writes?

A prompt that says "ask first" is not the same as an enforced permission boundary.

3. Can retries or timeouts create duplicate actions?

Think duplicate messages, bookings, invoices, records, or payments.

4. Is there a spend or usage ceiling that stops execution?

An alert after the cost is incurred is not an execution boundary.

5. Does every consequential action leave a readable decision trace?

The useful chain is context, tool, permission, approval, cost, outcome, and rollback reference.

6. Can the team stop, replay, or recover the workflow after a failure?

Recovery ownership and evidence matter more than a generic retry button.

Your control read

Run the A$99 risk audit See the complete sample

The scorecard is a screening tool, not a security certification or revenue guarantee. The paid audit reviews one real workflow and returns the top three risks plus a 14-day fix plan.