Security Research

Deep dives into real-world Agent Skill attacks and defenses.

Supply Chain2026-03-23

SANDWORM_MODE: The MCP Worm That Spreads Through npm Typosquats

19 typosquatted npm packages targeting Claude Code, Cursor, and Windsurf. Injects malicious MCP configs, steals credentials, and self-propagates through Git repos.

Threat Analysis2026-03-22

A Fake Polymarket Skill Stole Solana Wallets: Inside the bob-p2p Attack

How a ClawHub skill posing as a decentralized API marketplace tricked AI agents into storing private keys in plaintext and purchasing scam tokens.

Malware Analysis2026-03-22

One Actor, 677 Malicious Skills: The AMOS Stealer ClawHub Campaign

Deep dive into how a single threat actor weaponized ClawHub to distribute Atomic macOS Stealer to 7,000+ victims in 3 days.

MCP Security2026-03-22

Your MCP Server Is Reading Your SSH Keys: Tool Poisoning in Practice

Reproducing Invariant Labs' proof-of-concept: how a malicious MCP tool description silently exfiltrates credentials from Claude Desktop and Cursor.