relay-fleet

Warn

Audited by Gen Agent Trust Hub on May 24, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/relay-fleet.js uses child_process.spawn to manage parallel execution of child tasks. It dynamically constructs argument arrays that are passed to new Node.js processes.
  • [REMOTE_CODE_EXECUTION]: The script includes a --dispatch-script flag that allows the user to specify an arbitrary path for the Node.js file to be executed. This capability can be exploited to run malicious code if an attacker influences the agent's command-line arguments.
  • [REMOTE_CODE_EXECUTION]: Data from the --leaves-file (a JSON file) is parsed and its fields (such as test_command, branch, and executor) are passed as arguments to subprocesses. This provides an indirect injection surface where malicious data in the JSON file could influence the behavior of the child processes.
  • [DATA_EXFILTRATION]: The skill uses fs.readFileSync to load content from a path provided via the --leaves-file argument. This could be abused to read sensitive local files if the agent is directed to an unauthorized path.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 24, 2026, 03:00 AM
Security Audit — agent-trust-hub — relay-fleet