agent-design-review
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides structured instructions and templates for agent design and review. It emphasizes a security-first approach by advising users to enforce trust boundaries and approvals in code rather than relying solely on prompt instructions.
- [EXTERNAL_DOWNLOADS]: The skill references architectural guidance and technical documentation from trusted organizations, including Anthropic, OpenAI, and Google. These external resources are used neutrally to provide canonical implementation patterns and provider-specific configurations.
- [DATA_EXFILTRATION]: No sensitive file access, network exfiltration patterns, or hardcoded credentials were found. The skill explicitly guides developers to manage side effects and high-risk actions through approvals and deterministic policy gates.
- [INDIRECT_PROMPT_INJECTION]: While the skill defines templates that process user and retrieved content, it includes strong mitigation guidance. It recommends using XML tag delimiters and explicit decision policies to separate untrusted input from system instructions, and reinforces that security boundaries should be enforced at the runtime level.
Audit Metadata