delegate-multi-agent

Pass

Audited by Gen Agent Trust Hub on Mar 1, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The delegateChain function in delegate.js is vulnerable to indirect prompt injection when passing results between agents. \n
  • Ingestion points: Untrusted data enters the context through the context variable in delegateChain, which stores the output of the preceding agent. \n
  • Boundary markers: None. The agent output is concatenated directly into the prompt without delimiters or instructions to ignore embedded commands. \n
  • Capability inventory: The framework specifies that agents (e.g., 'coding' agent) may have high-privilege capabilities such as file_read, file_write, and file_edit. \n
  • Sanitization: There is no evidence of sanitization or escaping of the agent output before it is interpolated into subsequent prompts. \n- [COMMAND_EXECUTION]: The child_process module is imported and exec is promisified at the top of delegate.js, but these functions are not called within the script's logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 1, 2026, 05:11 AM