orchestrating-multi-agent-systems

Pass

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill facilitates multi-agent workflows that ingest and process user-supplied queries across various agent roles, creating a surface for indirect prompt injection.\n
  • Ingestion points: User queries enter the system via the execute method in assets/example_coordinator.ts and are handled according to the example_workflow.json configuration.\n
  • Boundary markers: The agent templates and system prompts do not utilize delimiters (e.g., XML tags or block quotes) or specific instructions to isolate or ignore potentially malicious commands embedded within user data.\n
  • Capability inventory: The skill is granted permissions for Write, Edit, and Bash(npm:*), providing a significant impact path if an agent's logic is subverted by a malicious payload in the processed data.\n
  • Sanitization: There is no evidence of input validation, filtering, or escaping of user-provided content before it is interpolated into agent contexts or used to determine routing logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 12, 2026, 12:03 AM