openclaw-zero-token

Fail

Audited by Gen Agent Trust Hub on Sep 12, 2026

Risk Level: HIGHCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads its full implementation and dependencies from an untrusted GitHub repository (github.com/linuxhsj/openclaw-zero-token.git).
  • [REMOTE_CODE_EXECUTION]: Fetches and executes a setup script from NodeSource's official domain to configure the Node.js environment.
  • [PRIVILEGE_ESCALATION]: The installation instructions require the user to run commands with root privileges (sudo) to install Node.js and system dependencies.
  • [COMMAND_EXECUTION]: Provides an exec tool that enables the AI agent to run arbitrary shell commands on the host machine.
  • [CREDENTIALS_UNSAFE]: The software captures and stores highly sensitive browser data, including session cookies and authentication tokens, in local files (data/auth/). This creates a significant risk of credential exposure if the untrusted codebase is malicious.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: External data enters the agent context through the web_fetch and web_search tools (SKILL.md).
  • Boundary markers: The documentation mentions prompt injection middleware, but there is no evidence of robust delimiters or warnings to ignore instructions in fetched data.
  • Capability inventory: The skill includes high-risk capabilities such as exec (shell execution), write (file system modification), and curl (network operations).
  • Sanitization: There is no description of sanitization or validation processes for data retrieved from untrusted web sources.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 12, 2026, 06:50 PM
Security Audit — agent-trust-hub — openclaw-zero-token