Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Agent Browser
v0.1.0Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking...
⭐ 0· 59·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description (browser automation) matches the content: CLI commands, snapshot/refs, session/state management, proxy, profiling and recording features. There are no unrelated required binaries or environment variables requested by the registry metadata. Templates and references are proportionate to the stated purpose.
Instruction Scope
SKILL.md and the reference docs instruct the agent to import browser sessions, save/restore plaintext state files, connect to a running Chrome via remote-debugging, run arbitrary JS (eval via base64/stdin), intercept/mock network requests, and use proxies. These are expected for a browser automation tool, but they are high-sensitivity actions (they can expose cookies, localStorage, CDP endpoints, and page contents). The templates also read environment variables (APP_USERNAME, APP_PASSWORD, AGENT_BROWSER_ENCRYPTION_KEY) even though no env vars were declared in metadata — this is optional usage but should be noted.
Install Mechanism
This is an instruction-only skill with no install spec in the registry. The README suggests normal installation routes (npm, brew, cargo) — no untrusted URLs, archive downloads, or hidden installers are present in the skill bundle itself.
Credentials
The registry lists no required credentials, which aligns with the skill. However the docs and templates reference environment variables (APP_USERNAME, APP_PASSWORD, AGENT_BROWSER_ENCRYPTION_KEY) and encourage using an 'auth vault' or env vars for credentials. Those env vars are optional for running the templates but are sensitive. No unrelated third-party keys or system-wide credentials are requested.
Persistence & Privilege
always:false (default) and the skill does not ask to be force-included or to change other skills' configs. It describes saving session/state files to disk (normal for this tool) but does not attempt to persist credentials into other skills or system-wide settings.
Assessment
This skill appears coherent for browser automation, but it legitimately performs high-sensitivity actions—read and follow these precautions before installing or running it:
- Treat state files (auth.json, auth-state.json, etc.) as secrets: do not commit them to version control and delete them when done. Use the recommended AGENT_BROWSER_ENCRYPTION_KEY if you persist state.
- Importing a running Chrome (--auto-connect) exposes all cookies and tokens from that browser to the tool; only do this on a trusted machine and close Chrome when finished.
- The tool supports eval of arbitrary JavaScript and exposing the CDP WebSocket URL; avoid running untrusted scripts or sharing CDP URLs as they can give full control over your browser session.
- Proxy/network-routing features can be used for legitimate geo-testing but can also route data through third-party proxies—ensure you trust proxy endpoints and avoid embedding credentials in plain text.
- The provided templates expect environment variables like APP_USERNAME/APP_PASSWORD even though they are not declared as required—set these in a secure environment (CI secrets, local protected env) and prefer the 'auth vault' workflow the docs recommend.
- Run automation that uses real credentials in an isolated environment (ephemeral profile or VM) where possible, and prefer short-lived sessions for CI.
If you want a stricter check before installing, ask the publisher for: (1) the official source/repository URL or release artifacts, (2) a signed binary or release on a well-known host, and (3) clarification which env vars are required vs optional. If you cannot verify the upstream source, run this tool only in isolated/test environments.Like a lobster shell, security has layers — review code before you run it.
latestvk974d41j5r3gxbhnbcate5dc1d83czz0
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
