subagent-arena
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists entirely of natural language instructions guiding the agent through a logical workflow for problem synthesis. No technical vulnerabilities or malicious patterns were detected.
- [PROMPT_INJECTION]: The skill processes user-defined charters, creating a potential attack surface for indirect prompt injection. Ingestion points: The skill takes user input to define charters for sub-agents in SKILL.md. Boundary markers: The instructions explicitly mandate that sub-agents be read-only and limited to investigation only. Capability inventory: The skill does not invoke shell commands, file writes, or network operations. Sanitization: No specific filtering is mentioned, but the workflow requires a synthesis step by the primary agent which acts as a verification layer.
Audit Metadata