Skip to content

DEV Community

← All Trends

AGENTS.md Poisoning and Security Risks

13 posts in this trend in the last 7 days • Active 1 day ago

mech.app

Oct 31

AGENTS.md Poisoning: Why Your Downloads Folder Is a Security Boundary for Coding Agents

#llm #security #agents #cybersecurity

5 min read

Oct 27

Markdown Is the New Executable: What the AGENTS.md-in-Downloads Attack Teaches Us About Coding Agent Security

#ai #security #agents #llm

7 min read

Oct 27

Beware of the Markdown File in Your Downloads Folder

#agents #ai #llm #security

4 min read

mech.app

Oct 27

Markdown as Malware: How AI Agents Turn Downloads Folders into Instruction Injection Vectors

#agents #ai #cybersecurity #security

6 min read

Dekita

Oct 30

Stop launching your coding agent from Downloads. Here is a 30-second pre-flight check.

#ai #agents #security #devtools

5 min read

Oct 30

Stop Untrusted Repo Files From Steering Coding Agents

#ai #security #agents #devops

9 min read

jamilxt

Oct 27

Stop Coding Agents From Trusting Downloads: Build a Workspace Preflight

#ai #security #python #agents

9 min read

Oct 28

Malicious Downloads Can Hijack AI Coding Agents and Steal Tokens

#agents #security #ai #cybersecurity

4 min read

mech.app

Oct 30

Indirect Prompt Injection via Repository Files: How Coding Agents Inherit Instructions from Untrusted Sources

#security #agents #llm #cybersecurity

5 min read

Oct 31

A web page told my coding agent what to do. It never asked me.

#ai #security #typescript #agents

3 min read

Oct 28

The AGENTS.md your agent reads as an order

#agents #ai #security #sandbox

3 min read

Sarah Reid

Oct 27

Your AGENTS.md Is an Attack Surface: How a Trusted File Hijacked Codex and Claude Code

#ai #security #agents #llm

5 min read

Oct 27

Prompt injection in AI agents: real attack patterns and the defenses that actually work

#ai #security #llm #agents

7 min read

Trend Analysis

Developers are discussing a newly uncovered security vulnerability where malicious instruction files like AGENTS.md placed in untrusted directories like Downloads hijack AI coding agents. This exploit leverages the inherent trust agents place in local workspace configuration files to execute unauthorized commands or install malicious packages.

Core Questions Under Debate:

How can developers prevent AI coding agents from trusting untrusted directories like Downloads?
What are the best practices for implementing preflight workspace security checks?
How do instruction file harnesses like AGENTS.md and CLAUDE.md handle malicious prompt injections?

First detected: October 31, 2026

Last updated: 1 day ago

Algorithm: Vector Leader Clustering (threshold: 82%)

About this trend

This trend groups posts from the last 7 days that discuss similar concepts, technologies, or ideas.