Skills
Navigate
skills/browser-use/terminal/browser-use-terminal/Gen Agent Trust Hub

browser-use-terminal

Pass

Audited by Gen Agent Trust Hub on Aug 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill's primary function is to execute arbitrary shell commands and Python scripts to automate browser tasks and perform repository work.
  • crates/browser-use-agent/src/tools/handlers/shell.rs implements shell and exec tools.
  • python/llm_browser_worker/worker.py uses exec() to run Python code snippets.
  • [EXTERNAL_DOWNLOADS]: The skill fetches its core components and dependencies from remote sources during setup.
  • scripts/install/install.sh and scripts/install-agent-ripgrep.sh download binaries from GitHub repositories.
  • [PROMPT_INJECTION]: The skill includes instructions to manage the agent's behavior and user interaction flow.
  • prompts/browser-tool-description.md instructs the agent to hide internal tool commands from the user to simplify the terminal interface.
  • [REMOTE_CODE_EXECUTION]: The setup instructions utilize a pattern of piping a remote script to a shell for installation from the official vendor domain.
  • AGENT_SETUP.md directs the user to run curl -fsSL https://browser-use.com/terminal/install.sh | sh.
  • [CREDENTIALS_UNSAFE]: The skill manages credentials for model providers and browser sessions.
  • crates/browser-use-llm/src/auth/codex.rs reads local configuration files like ~/.codex/auth.json to retrieve OAuth tokens.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 17, 2026, 04:52 AM
Security Audit — agent-trust-hub — browser-use-terminal