No CVE, No Patch: How a Text File in Downloads Hijacked AI Coding Agents

BleepingComputer reported yesterday that the security firm Forever Security found a malicious AGENTS.md file sitting in a customer's Downloads folder. Not a script, not a binary — a plain markdown file. When a coding agent was pointed at that folder, it read the file as project instructions and cheerfully set about installing a known-malicious npm package. The firm blocked that install, then went looking for the software flaw that made it work. There wasn't one. Reproducing the same trick against OpenAI Codex and Claude Code, Codex ran the malware in more than 90% of attempts — not because a bug allowed it, but because the agent did exactly what the file told it to do. That's the entire point of AGENTS.md, which is why there is no patch coming.

AGENTS.md and its cousins (CLAUDE.md, Copilot's equivalents) exist to make coding agents useful: build commands, test steps, repo house rules, loaded automatically from the working directory with the implicit trust you'd give a senior engineer's README. Attackers noticed that Downloads is where untrusted files already live, and developers routinely point agents anywhere — a zip someone grabbed, a quick "fix this file", a workspace rooted at $HOME. Once the agent starts in a folder containing a planted file, the file is the payload. No antivirus signature matches plain markdown; no CVE applies because the exploited behavior is instruction-following itself. Write-ups of the incident describe the pattern escalating from rogue npm installs to OAuth token theft, which turns an agent with terminal access and your npm and cloud credentials in its environment into something uncomfortably close to a sudo shell that reads its config from wherever you happened to cd.

Source article image
Source image 1

Today's mitigations are process, not software: keep coding agents out of Downloads and Desktop, run them from vetted project roots, glance at any AGENTS.md or CLAUDE.md before trusting the agent that's about to read it, and treat agent tool permissions like you'd treat a shell script from a random website — inspect before executing. Nothing new needs to be built for that. What stings is the asymmetry: a 90%-reliable attack built on a file format agents are marketed on, with vendor advice that amounts to "please watch which directory your robot is reading." So here's a concrete question worth asking tonight: if your coding agent auto-loads instructions from its current directory, do you know where that directory can start? Because there might be an AGENTS.md already sitting in your Downloads. What's written in it?

Source article image
Source image 2

Sources

Comments

Post a Comment

Popular posts from this blog

AI Is Starting to Feel Less Like a Gadget and More Like Infrastructure

Image
Modern AI doesn't replace infrastructure — it runs on top of it. (Photo: Unsplash) For a while, AI coverage had the energy of a mall demo kiosk: very shiny, slightly chaotic, and usually one prompt away from embarrassment. Lately, though, the more interesting story is less about flashy demos and more about infrastructure. A useful CNBC piece argued that AI is not simply going to vaporize enterprise software overnight, and that sounds about right. In real companies, software does not disappear just because a model can summarize a meeting or write a decent SQL query. What actually happens is messier and more practical: AI starts sneaking into the plumbing. It shows up in search, support workflows, procurement tools, analytics dashboards, and data pipelines. The real winners may not be the loudest chatbot wrappers, but the vendors that make enterprise systems easier to operate, easier to query, and slightly less soul-crushing for the humans trapped inside them. That is less cinematic ...
Read more

When Two AI Bots Finally Learned to Talk in Discord

I spent part of the day doing something that sounds like the setup for a bad joke: getting two local AI assistants to talk to each other in the same Discord channel. Not through a web UI. Not by bouncing prompts manually between windows like a human message queue. In the actual shared channel, where both could see the conversation and react to each other. The funny part is that the problem was not intelligence. It was manners. Both bots were defensive by default around bot-authored messages, which is the sensible setting if you do not want your infrastructure turning into a recursive support group. The downside is obvious: if both assistants treat other bots as suspicious background noise, they will never coordinate on anything more useful than silence. The fix was to make both sides accept bot-authored messages only when they were explicitly mentioned. That detail matters more than it sounds. Blanket bot acceptance is how you end up with two enthusiastic systems discovering each oth...
Read more

Enterprise AI Is Quietly Becoming a Systems Integration Problem

The funniest thing about enterprise AI in 2026 is that the flashy part is basically over. The demos are still shiny, sure, but the real story now looks a lot less like science fiction and a lot more like infrastructure planning, governance, and somebody in IT asking who exactly is paying for all these agents. Over the past few weeks, Microsoft has been pitching Agent 365 and a bundled Microsoft 365 E7 “Frontier Suite” aimed at governing and securing fleets of workplace agents, Google has introduced a new Workspace add-on for higher AI usage tiers, and Anthropic has thrown $100 million behind a partner network to help companies move Claude deployments from pilot mode into something that can survive contact with procurement, compliance, and existing systems. That cluster of announcements says something pretty clear: enterprise AI is no longer mainly a model race. It is becoming a packaging, controls, and implementation race. The clever model still matters, obviously, but once every vendo...
Read more