fleet-maintenance
Warn
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: MEDIUMPERSISTENCECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PERSISTENCE]: The skill modifies shell startup files (~/.profile and ~/.zprofile) to export environment variables and modify the PATH, ensuring specific configurations persist across shell sessions.
- [COMMAND_EXECUTION]: The skill makes extensive use of system commands such as dscl, fdesetup, sysadminctl, and codesign to audit and manage user accounts, security settings, and binary integrity.
- [REMOTE_CODE_EXECUTION]: The skill implements remote management by piping local scripts directly into interpreters on remote hosts via SSH (e.g., ssh HOST 'node -' < script.mjs), allowing for the execution of code on the target machines.
- [DYNAMIC_EXECUTION]: Several scripts utilize dynamic execution patterns, including node -e for inline JavaScript evaluation and perl -e for wrapping commands with timeout logic.
- [INDIRECT_PROMPT_INJECTION]: The skill relies on an external configuration file (inventory.json) to define the desired system state. If this file were modified by an unauthorized process, it could lead to the agent applying malicious system configurations.
- [EXTERNAL_DOWNLOADS]: The skill interacts with the external service https://octopool.openclaw.ai to manage GitHub CLI authentication and caching.
Audit Metadata