Markdown skill files propagate jailbreaks into AI agents
A poisoned SKILL.md raised mean attack success from 0.5% to 45% across seven open-weight models.
Stories per day · 30 days
Relatedprompt-injection2agent-skills1claude-code1codex1coding-agents1credential-theft1discord1file-sharing1jailbreak1langchain1nitro1
2 stories, newest first · one card per story
Stories tagged #markdown
Malicious AGENTS.md tried to hijack coding agents
Forever Security says a planted markdown file on a customer PC could push Codex to install malware or leak tokens, and was stopped before code ran.
1