ag2-middleware

Pass

Audited by Gen Agent Trust Hub on Jul 4, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The middleware system introduces surfaces for indirect prompt injection by allowing interception and modification of untrusted data.
  • Ingestion points: The on_turn, on_llm_call, on_tool_execution, and on_human_input hooks in SKILL.md process external data from LLMs and human users.
  • Boundary markers: The provided examples do not demonstrate the use of delimiters or 'ignore' instructions when processing intercepted content.
  • Capability inventory: The framework supports logging, retries, history trimming, and arbitrary mutation of tool arguments or prompts.
  • Sanitization: The documentation does not specify default sanitization or validation routines for content modified via middleware hooks.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 4, 2026, 08:34 AM
Security Audit — agent-trust-hub — ag2-middleware