orchestrating-agent-relay

Pass

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to use the agent-relay CLI tool for operational tasks, including starting a detached broker process, spawning worker agents, and managing their lifecycle through shell commands.
  • [EXTERNAL_DOWNLOADS]: The instructions guide the agent to install the agent-relay package from the public npm registry using standard package management commands.
  • [PROMPT_INJECTION]: The skill includes an orchestrator instruction template that uses directive language to define the agent's persona and operational protocol when managing the relay infrastructure.
  • [PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by instructing the orchestrator to read and process potentially untrusted text responses from worker agents via the agent-relay replies command.
  • Ingestion points: Worker replies are read from the agent-relay CLI output.
  • Boundary markers: No specific delimiters are suggested for the worker output.
  • Capability inventory: The orchestrator can spawn new agents and execute shell commands based on the content of worker replies.
  • Sanitization: No input validation or sanitization of the worker replies is described.
Audit Metadata
Risk Level
SAFE
Analyzed
May 19, 2026, 12:20 PM
Security Audit — agent-trust-hub — orchestrating-agent-relay