---
format: "aidr-story-markdown/v1"
id: "6d49b8b2d58349729f9d1e5fc634a3c5b7f225a4cb64188dfe5a5741bdd3ea65"
canonical_url: "https://aidr.today/6d49b8b2?lang=en"
title: "Git Exploit Runs Code Before Trust Prompt in 7 AI Agents, Leaving 4 Unpatched"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-02T17:02:45.000Z"
category: "Research"
topics: ["agent","multi-agent","security","coding"]
source_urls: []
summary: "A security flaw allows malicious Git repositories to execute attacker code before an AI assistant requests user permission. Four of the seven affected tools re…"
---

# Git Exploit Runs Code Before Trust Prompt in 7 AI Agents, Leaving 4 Unpatched

> [Open the canonical story](<https://aidr.today/6d49b8b2?lang=en>)

**Published:** 2026-09-02T17:02:45.000Z
**Category:** Research
**Topics:** agent, multi\-agent, security, coding

## Summary

A security flaw allows malicious Git repositories to execute attacker code before an AI assistant requests user permission\. Four of the seven affected tools re…

## Sources

_No safe source URL was recorded._