Undetectable AI

Security checks across malware telemetry and agentic risk

Overview

The code appears to be a local text tool, but the skill is explicitly designed to help AI-generated writing evade detectors and pass as human-written.

Avoid installing this skill if you need policy-compliant or transparent writing assistance. The reviewed code does not show data theft or hidden network behavior, but the skill’s stated purpose is to deceive AI-detection systems and downstream reviewers.

VirusTotal

55/55 vendors flagged this skill as clean.

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.

#
ASI09: Human-Agent Trust Exploitation
High
What this means

Using this skill could mislead teachers, reviewers, platforms, or detection systems about whether text was AI-generated and may violate submission or platform rules.

Why it was flagged

The skill directly states that it is intended to evade detection systems and make AI-generated text appear human for submission.

Skill content
Use to bypass AI detection, evade checkers, or humanize text before submission.
Recommendation

Do not approve or use this skill for evasion or undisclosed submission. If the goal is legitimate editing, reframe it as transparent style cleanup and remove bypass/evasion instructions.

#
ASI02: Tool Misuse and Exploitation
Low
What this means

The agent can run the CLI on local files and write transformed outputs, including batch processing if the user follows the examples.

Why it was flagged

The skill requests local file and shell capabilities. These are expected for a CLI that reads and rewrites documents, but they allow processing and creating modified files.

Skill content
allowed-tools:
  - Read
  - Write
  - StrReplace
  - Shell
  - Glob
Recommendation

Run it only on intended files, avoid sensitive documents, and review all output before sharing or submitting it.

#
ASI04: Agentic Supply Chain Vulnerabilities
Low
What this means

Installation may fail or pull packages from npm depending on how it is run.

Why it was flagged

The skill asks users to install npm dependencies before use. A package-lock is provided, but dependency installation still introduces supply-chain exposure, and the command appears to run from scripts while package.json is at the package root.

Skill content
cd scripts && npm install
Recommendation

Install only after inspecting the package files, prefer the checked-in lockfile, and correct the setup instructions to run from the directory containing package.json.