multi-agent
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOW
Full Analysis
- SAFE (INFO): No security issues or malicious patterns were identified in the provided skill files. The included Python script performs standard filesystem and configuration validation using secure methods.
- Indirect Prompt Injection (LOW): The skill defines a blueprint for multi-agent systems which naturally ingest untrusted data via the 'task' parameter. While this architecture has an inherent attack surface, the skill itself does not implement exploitable logic. 1. Ingestion points: 'task' parameter in SKILL.md. 2. Boundary markers: Absent in the parameter schema. 3. Capability inventory: Orchestrates agents that are documented to use 'web_search' and 'execute_code' tools. 4. Sanitization: No input sanitization is implemented in the structural validator.
Audit Metadata