Browser Automation V2

WarnAudited by ClawScan on May 10, 2026.

Overview

The skill matches its browser-automation purpose, but it builds shell commands from user-provided URLs, profile names, and form/search text in a way that could run unintended commands.

Install only if you trust the skill and can use it with a dedicated browser profile. Avoid passing untrusted URLs, search terms, form values, or profile names until the scripts are changed to avoid shell-based command construction.

Findings (4)

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.

What this means

A crafted URL, search query, form value, or profile name could cause the local shell to run commands the user did not intend.

Why it was flagged

The central command runner uses child_process.exec, which invokes a shell. Other methods build cmd strings from user-controlled values such as URL, text, key, ref, format, timeout, and BROWSER_PROFILE, creating command-injection risk.

Skill content
exec(cmd, { timeout }, async (err, stdout, stderr) => {
Recommendation

Replace exec string commands with execFile/spawn using an argument array, validate allowed values, and escape or reject shell metacharacters in all user-controlled inputs.

What this means

If an agent or user passes a maliciously crafted URL, the skill could misuse the local command shell instead of only fetching the intended page.

Why it was flagged

A user-supplied URL is inserted directly into a shell command. This exposes the OpenClaw CLI through an unsafe command-construction pattern.

Skill content
exec(`openclaw web_fetch "${url}" --max-chars 10000`, { timeout: 15000 }, (err, stdout, stderr) => {
Recommendation

Invoke web_fetch without a shell, pass the URL as a separate argument, and restrict accepted URL schemes to expected web URLs.

What this means

The automation may act through a browser profile that is already logged in, so page access, screenshots, PDFs, or form filling could involve private accounts.

Why it was flagged

The skill relies on a browser profile, which can contain authenticated session state, but the metadata declares no primary credential or config path and the artifacts do not clearly bound what logged-in sites or session data may be used.

Skill content
Browser profile configured (default: `openclaw`)
Recommendation

Use a dedicated low-privilege browser profile, document that logged-in sessions may be used, and require explicit user confirmation before interacting with authenticated or sensitive sites.

What this means

Users may not realize from metadata alone that the skill needs Node/OpenClaw browser CLI access, a gateway, and a browser profile.

Why it was flagged

The runtime requirements are documented in SKILL.md, but the registry metadata lists no required binaries, environment variables, credentials, or config paths. This is an under-declaration rather than hidden behavior.

Skill content
- OpenClaw v2026.2.15+
- Browser profile configured (default: `openclaw`)
- Gateway running
Recommendation

Declare runtime binaries, optional environment variables, and browser-profile requirements in metadata so users can assess permissions before installation.