self-improving-systems
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [SAFE]: The skill is entirely composed of markdown documentation and design patterns for architectural guidance. It contains no executable scripts, binaries, or active code components.- [INDIRECT_PROMPT_INJECTION]: The skill identifies memory as a major attack surface and provides proactive, detailed mitigations for indirect prompt injection and memory poisoning, specifically referencing safety research like the MINJA attack.- [EXTERNAL_DOWNLOADS]: All external tools referenced (such as mem0, Letta, and MemGPT) are well-known frameworks in the AI agent ecosystem. The plugin installation instructions correctly point to the author's specified marketplace and follow standard platform extension procedures.- [CREDENTIALS_SAFE]: No hardcoded credentials or sensitive file paths were detected. The skill advises on privacy and compliance requirements (GDPR/HIPAA) for state management.
Audit Metadata