Skip to content
Submit incident
AII-1000-56323
Well documentedThree or more independent publishers report the core facts.

Zero-click RCE vulnerability in AI coding agents Plugin4Shell

Researchers at cybersecurity startup AIR discovered a zero-click remote code execution vulnerability called Plugin4Shell affecting four major AI coding agents: Codex, Claude Code, Gemini CLI, and GitHub Copilot. The flaw allows attackers to execute malicious code by substituting a trusted plugin with malicious code, potentially giving attackers access to enterprise systems. Most vendors have released patches for the vulnerability.

18 September 2026 Β· OpenAI Β· Security and cyberattacks
18 September 2026
AII-1000-563237 sources ↓
Quick facts
Category
Security and cyberattacks
Country
United States
Source
TheRegister.com
Date
18 September 2026
Drafted by
Raidu Incident Detection

What happened

Researchers at cybersecurity startup AIR discovered a zero-click remote code execution (RCE) vulnerability affecting AI coding agents including OpenAI's Codex, Anthropic's Claude Code, Google's Gemini CLI, and Microsoft-owned GitHub Copilot. The vulnerability, named Plugin4Shell, allowed attackers to execute malicious code without developer interaction by swapping a trusted plugin from an online marketplace for a malicious one.

The flaw exploits how these agents handle plugin installations. When developers install a plugin, the agent downloads code from a Git repository and uses a secure hash algorithm (SHA) to verify it is an approved version. However, Claude Code, Codex, and GitHub Copilot pass the SHA directly to Git without subsequently verifying that Git actually checked out the commit corresponding to that SHA. An attacker controlling the plugin repository can create a malicious version and use the SHA of a legitimate commit as its name, causing the agent to execute malicious code instead of the trusted plugin.

Reported impact

The vulnerability could give an attacker the same reach into a company's systems and data as the employee running the agent, potentially exposing enterprise development environments. It was described as the first supply chain vulnerability of the AI agent ecosystem.

Index analysis

Most vendors released patches for Plugin4Shell after AIR reported the flaw. The researchers noted it is a flaw no marketplace can fix, requiring users to update their agent.

Classification

Category
Security and cyberattacks
Industry
Digital security
Country
United States
Harm status
Potential harm
Intent
Unknown
Timing
Post-deployment
Harm types
Public interest
Harmed
Business
Failure modes
Agent misaction
AI systems
Codex, Claude Code, Gemini CLI, GitHub Copilot

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

TheRegister.com
AI coding agents' 0-click RCE flaw could hand attackers keys to the kingdom
17 September 2026
Startup Fortune
Plugin4Shell Flaw Hits Claude Code, Codex, Copilot and Gemini CLI - Startup Fortune
18 September 2026
InfoWorld
A zero-click RCE flaw in AI coding agents could have exposed enterprise systems
18 September 2026
IT Security News - cybersecurity, infosecurity news
Zero-click RCE vulnerability hit four major AI coding agents, two remain unpatched - IT Security News
18 September 2026
SC Media
Bacteria, Spartans, AI gone wild, Cisco, WordPress, Settra, Plugin4Shell, Josh Marpet - SWN #617
18 September 2026
WION
One flaw hit all four major AI coding assistants! Two are still unpatched
20 September 2026
The Next Web
Plugin4Shell: the safety lock that never checked itself
21 September 2026