One Markdown File Can Hijack Your AI Coding Agent: The Hidden Cost of Blindly Trusting AGENTS.md
𝗝𝗼𝗵𝗻
- Comments13
𝗝𝗼𝗵𝗻
How a single AGENTS.md or CLAUDE.md file in your Downloads folder or a pull request can silently hijack your AI coding agent, steal OAuth tokens, and execute shell commands—plus how to lock down your local workflow.
If you run AI coding agents like Claude Code, OpenAI Codex, or Cursor CLI on your local machine, a single plain-text Markdown file is enough to compromise your environment. No memory corruption bug, zero-day exploit, or CVE is required.
In late October 2026, security researchers at Forever Security and the Aviatrix Threat Research Center documented the first in-the-wild AI coding agent hijacks triggered entirely by local AGENTS.md and CLAUDE.md files. Under controlled testing, the attack chain achieved up to a 90% success rate against Codex and 50% against Claude Code—redirecting API traffic, leaking OAuth session tokens, and executing arbitrary…
Excerpt from DEV Community ↗, by 𝗝𝗼𝗵𝗻.
One Markdown File Can Hijack Your AI Coding Agent: The Hidden Cost of Blindly Trusting AGENTS.md was first seen on DEV Community · Top this week on Nov 6, 2026. Since then it has appeared in 57 checks on 3 different days. Its best position was #20 of 30, reached on Nov 8, 12:43 UTC. It is #20 right now, climbing from #21 to #20 over the last 6 checks.
Position on DEV Community · Top this week at every check (lower is better).
Best rank per day (UTC).
♥ as reported by DEV Community.
Archived daily, weekly and monthly lists that include this entry.
Yes. As of Nov 8, 14:43 UTC it is #20, climbing from #21 to #20 over the last 6 checks.
#20 on DEV Community · Top this week, on Nov 8, 12:43 UTC. Trendoday checks this list every 60 minutes.
It was first seen on Nov 6, 2026 and has been on the list on 3 days (57 checks) since.