browser-use
Warn
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill provides commands that access and potentially exfiltrate sensitive information. The
cookies getcommand exports browser cookies, and theuploadcommand can be used to send local files to remote websites.- [COMMAND_EXECUTION]: The skill relies on theBashtool to executebrowser-useCLI commands for all browser operations.- [EXTERNAL_DOWNLOADS]: Theuvx browser-use installcommand triggers the download and installation of the Chromium browser binary from external remote sources.- [REMOTE_CODE_EXECUTION]: The skill usesuvxto fetch and execute thebrowser-usepackage from a remote package registry at runtime.- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted content from the web. Ingestion points includeuvx browser-use stateanduvx browser-use extract. There are no explicit boundary markers or instructions to isolate web content from the agent's command processing. The agent has significant capabilities including shell execution, file reading, and cookie management, which could be abused if malicious web content influences its behavior. No evidence of sanitization was found.- [DYNAMIC_EXECUTION]: Theuvx browser-use evalcommand allows for the execution of arbitrary JavaScript code within the context of a loaded web page, facilitating runtime code injection.
Audit Metadata