Drission Agent

ReviewAudited by ClawScan on May 10, 2026.

Overview

The package does not show data theft or destruction, but its claimed human-only safety wrapper is missing while it includes a local browser-control relay.

Review this skill carefully before installing. Do not set SOTA_NUCLEAR_CONFIRMED or start the relay on a personal browser session unless you are in an isolated sandbox. Ask the publisher to provide and review the missing secure_wrapper.py, requirements.txt, and referenced automation scripts before trusting the claimed security controls.

Findings (3)

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 user may trust safety controls that are not actually present in the reviewed package.

Why it was flagged

The provided package does not include secure_wrapper.py, and the included scripts use only an environment-variable gate, so the strong safety and human-interaction claims are not backed by the supplied artifacts.

Skill content
"secure_wrapper.py": **The ONLY entry point.** ... "AI Agents are physically unable to run any part of this toolkit without human interaction."
Recommendation

Do not rely on the claimed wrapper or human challenge unless the missing file is supplied and reviewed; treat direct script execution as possible whenever the gate environment variable is set.

What this means

The package cannot be verified as described, and users may need to obtain unreviewed files to make the documented workflow work.

Why it was flagged

SKILL.md references requirements.txt, secure_wrapper.py, force_takeover.py, and ultra_experiment.py, but those files are absent from the supplied manifest, leaving the declared install path, safety wrapper, and capabilities incomplete.

Skill content
4 file(s): SKILL.md, _meta.json, scripts/main_engine.py, scripts/python_relay.py
Recommendation

Ask the publisher to include all referenced files, pin dependencies, and ensure the reviewed manifest matches the documented entry points and capabilities.

What this means

If started, local processes or agents that can connect to the relay could interact with the browser automation endpoint during the relay lifetime.

Why it was flagged

The script creates a local unauthenticated TCP bridge from port 9223 to 9222, commonly associated with browser DevTools control. It is local and time-limited, but this high-impact browser-control path lacks the claimed wrapper challenge in the provided artifacts.

Skill content
server.bind(('127.0.0.1', local_port)) ... target_sock = socket.create_connection(('127.0.0.1', remote_port), timeout=5) ... SecureRelay().start(9223, 9222)
Recommendation

Run the relay only in an isolated sandbox, keep Chrome remote-debugging sessions separate from personal browsing, and require an explicit reviewed approval flow before starting it.