multi-agent
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a secure orchestration model using the Dispatcher-Planner-Worker-Verifier pattern, which isolates execution environments and prevents cross-agent context bleed.
- [COMMAND_EXECUTION]: Instructions include execution of local scripts for skill installation and workflow management. These operations are consistent with the framework's operational requirements and do not involve untrusted remote sources.
- [INDIRECT_PROMPT_INJECTION]: The orchestration model involves a Planner agent reading output from Worker agents. 1. Ingestion points: Worker handoffs and plan.json files. 2. Boundary markers: Structured JSON is used to delimit agent outputs. 3. Capability inventory: Execution of local scripts and sub-agent spawning. 4. Sanitization: No explicit sanitization of handoff text is described, though role isolation provides natural mitigation.
- [CREDENTIALS_UNSAFE]: The skill includes explicit pre-flight checks to ensure required API keys are configured in the environment without hardcoding them or exposing them in logs.
Audit Metadata