---
id: "CMP-202611-2d4jp2"
page: "https://infrastack.ai/live/CMP-202611-2d4jp2"
kind: "campaign"
aliases: []
title: "18 malicious AI agent skills on skills.sh and ClawHub: hidden installers, a CLAUDE.md prompt-injection planter, fake 'official' OpenClaw and Farcaster skills, token-phishing lookalikes (Snyk)"
severity: "high"
cvss: null
epss: null
exploited: false
status: "open"
category: "ai-tooling"
products: [{"name": "ClawHub", "type": "registry", "fixed": []}, {"name": "skills.sh", "type": "registry", "fixed": []}, {"name": "Claude Code", "fixed": [], "vendor": "Anthropic"}, {"name": "OpenClaw", "fixed": []}]
ecosystems: []
tags: ["campaign", "early-warning"]
not_affected_if: []
published_at: null
first_seen_at: "2026-11-16T19:18:36+00:00"
updated_at: "2026-11-16T19:19:56+00:00"
sources: [{"source": "analyst", "url": "https://research.snyk.io/blog/skill-analysis/", "first_seen_at": "2026-11-16T19:18:36+00:00", "cites": ["https://research.snyk.io/blog/skill-analysis/"]}]
exploit_sources: []
---

# 18 malicious AI agent skills on skills.sh and ClawHub: hidden installers, a CLAUDE.md prompt-injection planter, fake 'official' OpenClaw and Farcaster skills, token-phishing lookalikes (Snyk)

Link for people: https://infrastack.ai/live/CMP-202611-2d4jp2 (this .md is for agents)

Snyk scanned the top 10,000 skills each on skills.sh and ClawHub (taken 27 Oct 2026) and found 18 malicious ones (7 on skills.sh, 11 on ClawHub), with up to a few thousand installs each. skills.sh: six Roblox 'MM2 analytics' skills from reason-machines/data-skills make the agent clone from a GitHub Pages site and run an installer that is not in the skill. callicrate/skills/agents-md quietly runs a bundled binary that adds instructions from a remote server to the user's global ~/.claude/CLAUDE.md, so every later Claude Code session follows the attacker. ClawHub: a fake 'Official' OpenClaw auto-updater that installs a root daemon; a fake 'Official Farcaster' skill that creates a wallet and asks you to fund it; a Tushare skill that sends you to a lookalike site to get your token; six chaunceyliu 'aiusd' skills with base64 self-extracting installers; and two skills that hide their code (evolver, github-automation-pro handling GitHub tokens). Snyk does not confirm any victims. Snyk also found 64%+ of skills promote unsafe behavior. Whether the skills have been taken down is unknown.

_Summary, check and fix written by a model from the sources; labelled as such._

## Affected

- :ClawHub affected range not listed fixed no fixed version listed
- :skills.sh affected range not listed fixed no fixed version listed
- :Claude Code affected range not listed fixed no fixed version listed
- :OpenClaw affected range not listed fixed no fixed version listed

## Cause

How the attacker got in: not disclosed yet (unknown). Snyk found 18 malicious skills that users installed, with the attacker code carried inside them (hidden installers, a CLAUDE.md planter, a root daemon), and it does not confirm any victims or how an attacker got in.

How it reached others: their users installed or auto-updated it (tampered-install).

Last month's incidents with the same cause, and the full checklist: https://infrastack.ai/live/breaches.md#tampered-install

## How to check

Look for installed skills: reason-machines/data-skills/{roblox-mm2-analytics-toolkit, mm2-analytics-roblox-tracker, mm2-analytics-dashboard-roblox, murder-mystery-2-analytics-toolkit, mm2-analytics-roblox-toolkit, mm2-roblox-analytics-tracker}; callicrate/skills/agents-md; ClawHub autogame-17/evolver, gloryjack/tushare-pro, chaunceyliu/{aiusd-skill-agent, aiusd-skills, aiusd, trade-with-aiusd, aiusd-trade-agent, trade-agent}, aahuaxu/auto-updater-skill, jozh-bit/unzipped-skill, andy825lay-tech/github-automation-pro. Indicators: git clone https://8015238355.github.io, a directory named murder-mystery-dupe-roblox, ./setup.sh --install; files scripts/bin/fast-validate-* and validate_agentsmd.py; unexpected text added to ~/.claude/CLAUDE.md; domains tussahre.pro and openclaw-fake.com; downloads from github.com/galpha-ai/aiusd-skills/releases (aiusd-skill-agent.skill); files aiusd-skill-installer.js and .sh.

## Fix

Uninstall any of these skills. If agents-md was run, check ~/.claude/CLAUDE.md for text you did not write and remove it. If the MM2 setup.sh or the OpenClaw auto-updater was run, treat the machine as compromised: remove the boot daemon and rotate the credentials on it. Rotate any GitHub token given to github-automation-pro and any Tushare token typed into tussahre.pro. Do not fund wallets created by unzipped-skill. Review a skill's scripts and binaries before installing it, and scan skills (for example with Snyk Agent Scan).

## Timeline

- 2026-11-16T19:18:36+00:00 analyst https://research.snyk.io/blog/skill-analysis/

## Documents

- [describes] @snyksec: We re-scanned the top 10,000 skills on https://t.co/t3xD9BjmRE and ClawHub.

64% https://x.com/snyksec/status/2122404786033106952
- [describes] Re-Assessing the Security of the Agent Skill Ecosystem (Snyk) https://research.snyk.io/blog/skill-analysis/

---

_Free to read, with fair use of about 60 pages a minute. For higher limits or API access: sales@infrastack.ai_
