ag2-architect

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [SAFE]: No malicious patterns, exfiltration attempts, or obfuscation were detected in the skill instructions. The skill's behavior aligns with its stated purpose of assisting in multi-agent system design.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it processes user-provided descriptions to generate architectural designs and Python code.
  • Ingestion points: User-provided problem descriptions in SKILL.md.
  • Boundary markers: Absent.
  • Capability inventory: The skill generates functional Python code for multi-agent systems but does not execute it (SKILL.md).
  • Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 03:05 AM
Security Audit — agent-trust-hub — ag2-architect