connecting-agents-across-machines
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches markdown documentation from an external URL (
https://arelay.to/agent-relay) to check current state and confirm setup steps. - [COMMAND_EXECUTION]: The instructions involve executing
agent-relayCLI commands to manage workspaces, nodes, and agent spawning. It also includes instructions for global package installation vianpm install -g agent-relay. - [INDIRECT_PROMPT_INJECTION]: By establishing communication channels (DMs, threads, and shared inboxes) between agents on different machines, the skill creates a surface where an agent could receive and potentially act upon instructions sent by another agent.
- Ingestion points: Inter-agent messaging via
send_dmandpost_messagetools. - Boundary markers: No specific boundary markers or delimiters for message content are defined in the setup instructions.
- Capability inventory: The skill allows for spawning agents with specific tasks, executing commands on remote nodes, and writing to local files (e.g.,
received.txt). - Sanitization: No explicit sanitization or filtering of inter-agent messages is mentioned.
Audit Metadata