Back to feed
Dev.to
Dev.to
7/25/2026
MCP Security Audit: CVE-2026-30623, Tool Poisoning, and Config Hardening Checks

MCP Security Audit: CVE-2026-30623, Tool Poisoning, and Config Hardening Checks

Original: Model Context Protocol Through The Agent Stack Lens: What Broke, What's Fixed July 28, and What to Check Before Your Next mcp.json

Short summary

CVE-2026-30623 affects all four official MCP SDKs (Python, TypeScript, Java, Rust), allowing arbitrary shell command execution via unsanitized config values passed through STDIO transport. OX Security found 200k+ vulnerable instances across 150M+ downloads; Anthropic confirmed the behavior was intentional and declined to change it. The author provides a jq audit script for spotting at-risk STDIO servers and warns about tool-poisoning attacks hidden in tool description fields.

  • •CVE-2026-30623 lets config-file influence trigger arbitrary shell commands via MCP STDIO transport across all official SDKs
  • •Anthropic acknowledged the design decision but declined to fix it; 200k+ instances exposed
  • •Tool-poisoning attacks embed malicious instructions in tool descriptions that models follow without external detection

Generated with AI, which can make mistakes.

#ai-agents#ai-tools#open-source
Read full article at Dev.to

Is this a good recommendation for you?

Comments

Failed to load comments. Please try again.

Explore more

MCP Security: The Risks of Model Context Protocol and How to Govern It (2026)
Dev.to
MCP Servers face prompt injection
Dev.to
MCP Protocol Security: Why the readOnlyHint Vulnerability Exposes a Fundamental Flaw in AI Agent Tool Calls
Dev.to
The original title is: "MCP spec update: stateless sessions, deprecated features, and migration guide"
Dev.to