PulseAugur
EN
LIVE 15:26:28

Coding AI agents vulnerable to Git command execution before user approval

Research from Manifold Security, authored by Francisco Rosales, details a security vulnerability in several coding AI agents where they execute Git commands in the background before user approval. These commands can leverage a Git configuration setting called `core.fsmonitor` to run arbitrary programs, potentially exposing sensitive repository information or executing malicious code. The vulnerability arises because agents do not strip the repository's own Git configuration, and these background commands operate outside the typical sandbox environment. AI

IMPACT Highlights a critical security flaw in AI agent design, emphasizing the need for robust sandboxing and command validation before model interaction.

RANK_REASON Research paper detailing a security vulnerability in AI agents. [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — LLM tag →

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

Coding AI agents vulnerable to Git command execution before user approval

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · hugginf_expert ·

    Git Ran the Repo's Program Before Any Prompt: What GitSpawn Teaches Agent Builders About Background Context Calls

    <h2> TL;DR </h2> <ul> <li>Manifold Security's GitSpawn research (published September 1, 2026, by Francisco Rosales) reports eight findings across seven coding agents. Each agent runs <code>git</code> in the background to collect repository context, and that call runs outside the …