Browser Use Local
ReviewAudited by ClawScan on May 1, 2026.
Overview
This is a disclosed local browser-automation playbook with helper scripts, but it gives powerful browser and LLM-agent capabilities that users should scope carefully.
Install/use this only if you want local browser automation. Before running it, verify the browser-use environment and dependencies, use scoped LLM credentials, keep tasks specific, avoid sensitive logged-in sessions unless necessary, and delete screenshots, HTML dumps, QR crops, or persistent sessions that may contain private data.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
If invoked on the wrong site or with an overbroad task, the agent could click, type, inspect, or capture page content beyond what the user intended.
The skill intentionally exposes broad browser automation actions. This is expected for the stated purpose, but those actions can affect websites or accounts if used on logged-in sessions.
open pages, click/type, take screenshots, extract HTML/links
Use the skill only for specific, user-approved browser tasks and avoid letting it operate on sensitive logged-in accounts without clear instructions.
The configured API key authorizes calls to the selected LLM provider and may incur cost or expose tasks to that provider.
The bundled LLM-agent script reads an OpenAI-compatible API key and provider URL from the environment. This is disclosed and expected for the integration, with no hardcoded key or logging shown.
api_key = os.getenv("OPENAI_API_KEY")
base_url = os.getenv("OPENAI_BASE_URL")Use a scoped API key where possible, verify the OPENAI_BASE_URL, and avoid placing highly sensitive information in browser-agent tasks unless acceptable for that provider.
Users may need to rely on a preexisting local environment or install dependencies separately, which can introduce version or provenance uncertainty.
The artifacts reference browser-use, a virtualenv, and Python libraries, but do not provide a pinned installation specification. This limits setup/provenance clarity but does not show hidden installation behavior.
No install spec — this is an instruction-only skill.
Verify the browser-use installation and Python dependencies before use, preferably from trusted sources with pinned versions.
Browser state, screenshots, HTML dumps, and QR-code crops may contain sensitive session or login information and could be reused or left on disk.
The skill recommends persistent browser sessions, which can retain browsing state across steps. This is useful for the purpose but can also preserve sensitive page or login context.
Use persistent sessions to do multi-step flows: `--session <name>`.
Use separate session names for sensitive work, clean up screenshots/HTML/image outputs, and avoid persisting sessions longer than needed.
Tasks handled by the browser agent may be processed through the configured external LLM provider.
The skill includes an LLM-backed browser agent flow using a configurable provider endpoint. This is disclosed and purpose-aligned, but users should understand the provider boundary.
run an Agent with an OpenAI-compatible LLM (e.g. Moonshot/Kimi) using a custom base_url
Confirm the provider endpoint and data-handling expectations before using the LLM-agent workflow on sensitive pages or tasks.