User hacked via link from Claude chat, poisoned SKILL.md found in backup
skdh · x · 2026-08-29
User @Numalunah shared a real hacking story: they asked Claude in chat for a transcription app, and the download link Claude provided pointed to a copycat site bundling malware. Pasting the command into the terminal executed it instantly.
The scarier part came during backup restoration: a poisoned SKILL.md for Claude Code was hidden in the backup, looking exactly like the user's own writing style guide but containing instructions to silently re-download the malware and steal credentials every time the AI loaded it. Restoring that single file would have compromised the fresh laptop on day one. What saved the user: reading every skill, hook, and config file before letting the AI touch them. Lessons: AI assistants hand you links they never verified, and AI config files must be audited manually.
More from coding & agent
- Void v1.1.11 open sourced: one-command build and deploy to Cloudflare — cnakazawa · 2026-10-15
- Proof Launches x401, an Open Protocol Letting AI Agents Prove Who They Act For — csuwildcat · 2026-10-15
- Dev pushes back on "AI coding means insecure apps": humans always shipped bugs — dreamwieber · 2026-10-15
- Eight non-negotiable principles for building AI that actually learns from experience — tokenbender · 2026-10-15
- Animated teardown of XBOW's Chrome full-chain exploit that won Google's first $250K VRP bounty — moyix · 2026-10-15
- pstack's durable core: 10-20 line skills that distill agent best practices into principles — HankYeomans · 2026-10-15