llm-architect
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION] (SAFE): The skill instructions define a professional persona and do not contain any bypass attempts or 'ignore previous instruction' patterns.
- [CREDENTIALS_UNSAFE] (SAFE): Code examples use standard client initializations without hardcoded API keys or secrets.
- [DATA_EXFILTRATION] (SAFE): No patterns for accessing sensitive local files (~/.ssh, ~/.aws) or sending data to unauthorized external endpoints were detected.
- [EXTERNAL_DOWNLOADS] (SAFE): The skill does not perform runtime downloads of scripts or packages.
- [INDIRECT_PROMPT_INJECTION] (SAFE): While the skill assists in building RAG systems, it proactively includes validation rules (e.g.,
context-no-delimiter) specifically designed to mitigate injection risks in the applications it helps build. - [OBFUSCATION] (SAFE): All content is in clear text with no hidden characters, Base64 encoding, or homoglyph attacks.
Audit Metadata