install-openclaw
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The installation script uses subprocesses to execute various system commands for package management (
dnf,yum,npm), process inspection (ps,fuser,lsof), and service control (openclaw gateway). - [PRIVILEGE_ESCALATION]: Administrative privileges via
sudoare employed to install system-wide dependencies and global Node.js packages, which is standard practice for software installation on Linux environments. - [EXTERNAL_DOWNLOADS]: The skill fetches software and components from official system repositories and the Alibaba Cloud npm mirror (
npmmirror.com). These are well-known and reputable sources for the intended user base. - [DATA_EXFLLTRATION]: During the setup process, the script performs a network request to official Alibaba Cloud Model Studio endpoints (e.g.,
dashscope.aliyuncs.com) to validate the provided API key and configuration settings before final implementation. - [DYNAMIC_EXECUTION]: The script dynamically locates and executes a local installation script (
install.sh) when setting up the optional tokenless plugin, using paths found at runtime within the system's shared directories. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied parameters such as model IDs and billing plans through prompt interpolation. While the input is technically untrusted, the script mitigates risks by using safe subprocess handling (list-based execution) rather than raw shell strings.
- Ingestion points: Command-line arguments passed from the agent to
install_openclaw.pyas defined inSKILL.mdexamples. - Boundary markers: None present in the command templates.
- Capability inventory: The script possesses network access, file system write permissions for configuration, and the ability to execute system commands with elevated privileges.
- Sanitization: The implementation uses Python's
subprocess.runwith argument lists, which prevents shell metacharacter injection from user-provided strings.
Audit Metadata