Snyk scan of 20,000 agent skills identifies malicious CLAUDE.md injection

·2026.11.07 17:00

Key point

Snyk's analysis of 20,000 skills flagged over 65% for unsafe behavior but identified only nine as actively malicious, including one that appends remote text to global configuration.

Details

Snyk scanned 20,000 popular agent skills from skills.sh and ClawHub, flagging more than 65% for unsafe behavior while classifying only nine as actively malicious.
A skill named agents-md was reported to run a hidden binary that downloads remote text and appends it to ~/.claude/CLAUDE.md, allowing later Claude Code sessions to load the injected instructions.
Recommended mitigations include reviewing global instructions, verifying publishers, pinning reviewed MCP versions, and keeping wallet keys away from agents.
Pinning versions alone does not establish safety, and the article notes that Snyk also sells scanning services.

claude-code supply-chain-risk ai-security agent-skills snyk

Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.