iii-worker-manager
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill architecture describes a worker management system that ingests untrusted data from external SDK workers to be processed by an engine.
- Ingestion points:
AuthInput(headers, query parameters) andMiddlewareFunctionInput(payload) defined inSKILL.md. - Boundary markers: The documentation provides explicit instructions for using
rbacconfiguration blocks andexpose_functionsfilters to strictly limit the attack surface of connected workers. - Capability inventory: The skill demonstrates the use of
iii.triggerwithin middleware to forward requests and registration hooks to manage system triggers. - Sanitization: The reference middleware implementation demonstrates how to enrich payloads with verified caller identity (
_caller_id) and recommends using middleware for further request validation.
Audit Metadata