Agent Launchpad
Agent Launchpad guides an agent from proposed use to governed readiness. It coordinates the required records but does not override an incomplete gate.
Recommended sequence
Section titled “Recommended sequence”- Define the agent boundary, purpose, prohibited purposes and owners.
- Register the agent and runtime identity.
- Complete system intake, ACRS, ATF and relevant MAESTRO assessment.
- Establish data movement, tools, connections and least-privilege permissions.
- Draft and independently approve exact Runtime Access Policies.
- Test simulations, approval, denial, timeout, retry and containment paths.
- Resolve critical findings and accept residual risk through authorised governance.
- Enable only the intended environment and monitor the first operating period closely.
Readiness does not equal authority
Section titled “Readiness does not equal authority”Launchpad progress indicates that governance work has been completed or remains outstanding. Live execution still requires Gateway to validate authenticated identity, connection, permission, policy, approval and request context at runtime.
Reassessment
Section titled “Reassessment”Return to Launchpad after changes to purpose, owner, model, tools, data, supplier, environment, autonomy, delegation or risk. Suspend authority first when the change invalidates the trusted boundary.