HomeIntelligenceBrief
VULNERABILITY BRIEF 🔴 Critical Vulnerability

Zero‑click RCE (Plugin4Shell) compromises four AI coding agents, two remain unpatched

A zero‑click remote code execution flaw called Plugin4Shell bypasses SHA‑pinning in Claude Code, Codex, GitHub Copilot and Gemini CLI, letting attackers inject malicious code without user interaction. The issue highlights a gap in software‑supply‑chain controls that auditors and compliance teams must evidence.

Verisq™ Intelligence · 📅 September 18, 2026 · 📰 helpnetsecurity.com
🔴
Severity
Critical
VU
Type
Vulnerability
🎯
Confidence
High
🏢
Affected
2 sector(s)
Actions
3 recommended
📰
Source
helpnetsecurity.com

Zero‑click RCE (Plugin4Shell) compromises four AI coding agents, two remain unpatched

What Happened — A zero‑click remote code execution flaw, dubbed Plugin4Shell, bypasses SHA‑pinning in the plugin update mechanism of four leading AI coding agents (Claude Code, Codex, GitHub Copilot, Gemini CLI). The bug lets an attacker replace a trusted plugin with malicious code without any user interaction. Two of the four agents have not yet released a patch.

Why It Matters for Trust & Control Assurance

  • The scenario tests the control objective of software‑supply‑chain integrity – ensuring that third‑party components cannot be altered after review.
  • Continuous monitoring of plugin provenance and cryptographic verification provides the evidence needed for a defensible audit trail.
  • Verisq’s Third‑Party Risk Management capability can surface such supply‑chain gaps early and supply the documentation required for compliance frameworks.

Who Is Affected – Developers and enterprises that integrate AI coding assistants into their development pipelines; SaaS platforms that embed these agents; any organization relying on plugin marketplaces for AI tooling.

Recommended Actions

  • Inventory all AI coding agents and plugins in use; map them to your software‑supply‑chain control set.
  • Enable strict verification of plugin hashes and enforce policies that reject unsigned or mutable commits.
  • Prioritize patching the two unpatched agents; if patches are unavailable, consider temporary mitigation (e.g., disabling auto‑updates).
  • Capture evidence of these controls in a continuous‑monitoring repository to demonstrate audit readiness.

Technical Notes – The vulnerability exploits a flaw in how the agents fetch and verify pinned Git commits. By naming a branch with a 40‑character hexadecimal string, an attacker can cause the checkout process to accept a malicious commit while the SHA appears unchanged. The bug is zero‑click because the malicious code executes during the agent’s automatic background update, without any user action. Source: Help Net Security

📰 Original Source
https://www.helpnetsecurity.com/2026/09/18/plugin4shell-ai-coding-agents-vulnerability/

This Verisq Intelligence Brief is an independent analysis. Read the original reporting at the link above.

Vulnerability exposure

Is this CVE in your supply chain?

Verisq matches published vulnerabilities against the software your vendors run, so you know which relationships this touches before the next review.

See vendor exposure →