Knowledge Base Regulation Tracker Standards & Frameworks OSS Directory AI Incident Database Research Paper Library Case Studies Society Articles News Jobs About Charter Sign inBecome a Member
Home/AI Incident Database/Critical command injection in the mcp-remote client used to connect agents to remote MCP servers

Critical command injection in the mcp-remote client used to connect agents to remote MCP servers

A malicious or compromised remote MCP server could execute arbitrary operating system commands on any machine that connected to it through mcp-remote, a widely used bridge with hundreds of thousands of downloads.

Severity: HighMCPCommand injectionOAuth flowCVSS 9.6Client-side

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.

Submit incidentReport a correction or request removal

Related in AI Incident Database

Popular MCP server adds a hidden BCC to every email it sends after fifteen clean releases

An npm package that let AI agents send email through the Postmark service shipped fifteen legitimate versions before version 1.0.16 quietly copied every outgoing message to the developer's own server.

Supply chain compromise2025

Malicious prompt inserted into a released version of an AI coding extension instructs it to wipe systems

An attacker submitted a pull request that added a prompt telling the assistant to clean a system to a near-factory state and delete cloud resources; it shipped in an official release, and the attacker said the goal…

Supply chain compromise2025

Cryptomining payload injected into a widely used computer vision library through GitHub Actions cache poisoning

Attackers exploited a template injection weakness and cache poisoning in the project's GitHub Actions workflows to publish YOLO releases carrying a cryptominer, twice, showing that ML libraries are prime supply chain…

Supply chain compromise2024