Browser Use.Conflict

Security checks across malware telemetry and agentic risk

Overview

This is a browser automation skill, but it exposes broad browser-session, cookie, cloud/tunnel, and persistent code-execution capabilities that are not tightly scoped.

Install only if you trust the external browser-use CLI and need this level of browser control. Use an isolated test profile, avoid existing Chrome profiles and cookie export unless explicitly required, do not run Python/scripts from untrusted sources, be careful with cloud/tunnel commands, and close sessions when done.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Risk analysis

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.

#
ASI03: Identity and Privilege Abuse
High
What this means

The agent could act as the user on logged-in websites or expose reusable session cookies if those commands are used.

Why it was flagged

The skill documents using a real Chrome profile with existing authenticated sessions and provides cookie export/import commands, but the artifacts do not bound which accounts, cookies, or sites may be used.

Skill content
browser-use --profile "Default" open <url>      # Real Chrome with Default profile (existing logins/cookies)
browser-use cookies export <file>         # Export to JSON
Recommendation

Prefer a fresh test browser profile. Use --profile, --connect, and cookie export/import only when the user explicitly asks and understands the account/session impact.

#
ASI05: Unexpected Code Execution
Medium
What this means

If invoked carelessly, this could run local code with access beyond the browser task, depending on the external CLI's sandboxing.

Why it was flagged

The allowed browser-use command surface includes persistent Python execution and running local Python files, which is broader than simple browser navigation and extraction.

Skill content
browser-use python "code"                 # Execute Python (variables persist across calls)
browser-use python --file script.py       # Run file
Recommendation

Do not use the Python subcommand or run script files unless explicitly requested by the user and sourced from trusted, reviewed code.

#
ASI02: Tool Misuse and Exploitation
Medium
What this means

The agent could invoke provider API operations or other powerful browser-use subcommands outside the user's immediate expectation.

Why it was flagged

The wildcard tool permission exposes all browser-use subcommands, including raw cloud API passthrough, without artifact-level limits or confirmation requirements for higher-impact actions.

Skill content
allowed-tools: Bash(browser-use:*)
browser-use cloud v2 GET /browsers        # REST passthrough (v2 or v3)
browser-use cloud v2 POST /tasks '{"task":"...","url":"..."}'
Recommendation

Constrain use to the minimal browser commands needed for the task, and require explicit confirmation before cloud API passthrough, account/session mutation, uploads, or other high-impact actions.

#
ASI04: Agentic Supply Chain Vulnerabilities
Medium
What this means

Users may have difficulty verifying which publisher/version they are trusting, especially because the external CLI can access sensitive browser state.

Why it was flagged

The embedded metadata differs from the provided registry metadata, which lists a different owner ID, slug browser-use-conflict, and version 1.0.0. The skill also has unknown source/homepage and no install spec for the required CLI.

Skill content
"ownerId": "kn71fxj97n86164tdd84bymp3n7zypxq",
  "slug": "browser-use",
  "version": "2.0.0"
Recommendation

Verify the publisher and intended browser-use CLI source before installation, and prefer a package whose registry metadata, embedded metadata, and dependency provenance match.

#
ASI07: Insecure Inter-Agent Communication
Low
What this means

Browser traffic or local services may be exposed to a third-party cloud/tunnel provider if these features are used.

Why it was flagged

The skill documents cloud browser connections and Cloudflare tunnels, which can route browser activity or expose local services through external infrastructure.

Skill content
browser-use cloud connect                 # Provision cloud browser and connect
`cloud connect` provisions a cloud browser, connects via CDP, and prints a live URL.
browser-use tunnel <port>                 # Start Cloudflare tunnel (idempotent)
Recommendation

Use cloud and tunnel commands only when needed, avoid sensitive accounts or private local services, and close the connection when finished.

#
ASI10: Rogue Agents
Info
What this means

Open sessions may continue to exist until closed, which matters more if the browser is logged into real accounts.

Why it was flagged

The background daemon and persistent browser are disclosed and cleanup is documented, but users should notice that browser state can remain active across commands.

Skill content
A background daemon keeps the browser open across commands, giving ~50ms latency per call.
5. **Repeat**: browser stays open between commands
6. **Cleanup**: `browser-use close` when done
Recommendation

Run browser-use close when finished, and avoid persistent real-profile sessions unless they are necessary.