agent-safety
Pass
Audited by Gen Agent Trust Hub on Aug 6, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves exclusively as a defensive security resource, providing guidelines to bound agent autonomy rather than executing dangerous tasks.
- [SAFE]: Promotes the principle of least privilege by advocating for deny-by-default tool access and the use of short-lived, task-scoped credentials.
- [SAFE]: Includes Python code examples demonstrating security best practices, such as path traversal prevention and blacklisting sensitive file types like .env and private keys.
- [SAFE]: Provides specific mitigation strategies for Indirect Prompt Injection (LLM01), emphasizing the use of delimiters and the segregation of instruction and data channels.
- [SAFE]: Recommends robust runtime kill-switches and monitoring, including tool-call rate limits, per-session cost caps, and human-in-the-loop requirements for irreversible actions.
Audit Metadata