autonomous-multi-ai-agents
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates a multi-agent environment where agents ingest untrusted data from various external sources while maintaining capabilities to modify the local environment.
- Ingestion points: The skill ingests data from user inputs (both text and voice), project files, source code repositories, and shared conversation threads, as specified in the resources section of SKILL.md.
- Boundary markers: The instructions do not define technical delimiters or specific prompt isolation techniques to differentiate between trusted instructions and untrusted data.
- Capability inventory: The system is authorized to edit code, files, and folders, and manage voice input/output streams, providing a significant impact surface if agents are manipulated.
- Sanitization: The skill relies on high-level coordination strategies (edit turns, conflict resolution, and human review) but does not implement technical sanitization or escaping of external content before it is processed by the agent models.
Audit Metadata