agents-manager
Pass
Audited by Gen Agent Trust Hub on Aug 7, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a detailed reference guide and does not contain executable code or scripts that perform suspicious operations. It focuses on educating developers about agent architecture, memory management, and debugging.
- [PROMPT_INJECTION]: The documentation includes explicit guidance on defending against prompt injection attacks, recommending specific XML-based defensive patterns to ensure agents treat user input as data rather than instructions.
- [DATA_EXFILTRATION]: Instructions within the skill specifically warn against including secrets, credentials, or proprietary code in external queries, demonstrating a security-conscious approach to data handling.
- [COMMAND_EXECUTION]: While the skill discusses the use of the Bash tool for sub-agents, it stresses the importance of least-privilege access and tool auditing to prevent permission sprawl and unsafe autonomy.
- [SAFE]: References to local skill paths are consistent with the author's identity and provide configuration examples for the user's local development environment.
Audit Metadata