AGENTFORCE-AF-07: Agent Fabric Broker Unconstrained Autonomy
🟡 Medium · MuleSoft Agent Fabric
Detects Agent Fabric broker LLM nodes that pick freely among three or more actions, including writes or delegation to other agents, with no loop cap, no INPUT_REQUIRED hand-back to the user and no inbound policy on the broker, so one request can chain many consequential calls.
Details
| Field | Value |
|---|---|
| Rule ID | AGENTFORCE-AF-07 |
| Severity | Medium |
| Category | MuleSoft Agent Fabric |
Remediation
Constrain the broker: set reasoning.max_number_of_loops (10 or less), split consequential actions into deterministic executors behind an approval step (echo TASK_STATE_INPUT_REQUIRED + router), give each subagent only the actions its task needs, and bind an inbound policy (authentication and rate limiting) to brokers.<id>.interfaces.a2a.policies.inbound.