openclaw-browser-setup
Warn
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill instructions in
SKILL.mddirect the agent to retrieve authentication tokens usingopenclaw config get gateway.auth.token. This exposes sensitive credentials to the agent's context during troubleshooting. - [INDIRECT_PROMPT_INJECTION]: The skill captures external web content via
openclaw browser snapshotandscreenshotcommands inSKILL.md, creating a path for untrusted data to enter the agent's environment. -- Ingestion points: HTML snapshots and page screenshots from arbitrary URLs accessed for verification. -- Boundary markers: Absent. The instructions do not specify delimiters or security warnings to prevent the agent from interpreting instructions embedded in the captured web data. -- Capability inventory: The skill allows the agent to modify system configurations (openclaw config set), approve device access (openclaw devices approve), and restart services (openclaw gateway restart). -- Sanitization: Absent. No filtering or validation steps are provided for the content retrieved from external sources. - [COMMAND_EXECUTION]: The skill makes extensive use of the
openclawCLI for system-level actions including service restarts and device identity management.
Audit Metadata