Remote Experiment Execution
Warn
Audited by Gen Agent Trust Hub on Apr 19, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill is designed to execute arbitrary bash commands on a remote server using a specialized CLI tool.
- [REMOTE_CODE_EXECUTION]: The primary function is to synchronize local code and execute it on a remote compute node, which facilitates remote code execution.
- [AUTONOMY_RISK]: Core behavior rules ('Rule 1' and 'Rule 2') explicitly instruct the agent to suppress human-in-the-loop oversight by executing commands and performing self-repair loops without waiting for user confirmation.
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read and analyze output from remote commands to drive its next autonomous actions, creating a vulnerability where malicious output from a remote process could manipulate the agent's logic.
Audit Metadata