PulseAugur
EN
LIVE 06:37:18

Malicious websites exploit AI coding assistants via Downloads folder

By PulseAugur Editorial · [1 sources] ·

A security firm named Forever Security discovered a vulnerability where malicious websites can write a file named AGENTS.md to a user's Downloads folder without any interaction. This file can then be interpreted by AI coding assistants like Codex and Claude Code. Codex, when operating in a "trusted" Downloads folder, executed a malicious npm package installation in over 90% of tests. Claude Code, using a similar CLAUDE.md file, routed traffic through an attacker's server approximately 50% of the time, potentially exposing login tokens. The researchers noted that the issue lies with how these agents process local markdown files, rather than a flaw in OpenAI or Anthropic's models themselves. AI

IMPACT Highlights a critical security flaw in AI agent file processing, potentially impacting enterprise adoption of coding assistants.

RANK_REASON Security vulnerability discovered in AI coding assistants related to file handling.

Read on dev.to — Claude Code tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Malicious websites exploit AI coding assistants via Downloads folder

How we ranked this

Signal score
39 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Security vulnerability discovered in AI coding assistants related to file handling.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
safety, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — Claude Code tag TIER_1 English(EN) · Gabriel Anhaia ·

    A Web Page Dropped AGENTS.md in Downloads. Codex Obeyed It in 90%+ of Tests

    <ul> <li> <strong>Book:</strong> <a href="https://www.amazon.com/dp/B0HCC3G7T4" rel="noopener noreferrer">AI That Ships</a> </li> <li> <strong>The series:</strong> <em>AI in TypeScript</em> — 5 books, from your first LLM call to agents in production — <a href="https://xgabriel.co…

RELATED ENTITIES

RELATED TOPICS