Skip to content
NewA calmer, faster ZeroHour

MCP Explained walks through the 2025-11-25 protocol and attack cases

Original title: MCP Explained

Hacker News · securityby rwilinski ()3 min read1 · 1 read
Part of a story covered by 2 sourcesResearchers release dataset of 24 validated MCP vulnerabilitiesMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

The guide covers JSON-RPC shapes, deprecated batching, and defenses for tool poisoning, shadowing, and injection.

A Runlayer site, MCP Explained, walks through the Model Context Protocol specification dated 2025-11-25. It shows JSON-RPC requests, results, errors, and notifications, and marks batching and parts of sampling, transports, and authorization as deprecated. A separate section names attacks that are not in the spec, including tool poisoning, tool shadowing, token passthrough, prompt injection through tool results, and elicitation phishing. It also previews a 2026-07-28 spec and extensions such as Skills, Apps, and Server Cards.

Key facts
  • Every MCP message is described as JSON-RPC 2.0 over UTF-8 JSON.
  • Batching of message arrays was deprecated on 2025-06-18.
  • The 2026-07-28 draft would require resultType on every result.
  • The next release is planned for March 2027, according to the page.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

3 min read at mcp-explained.com

Authorization has deprecated partsOAuth 2. 1 for remote servers Skills over MCP beta extensionAgent Skills served by a server Spec 2026-07-28 ↗Stateless, no handshake, multi round-tripExtensions ↗Apps, Skills, Tasks, Server Cards, authSkills over MCP →Agent Skills as an MCP extension · lessonProgressive disclosure ↗Tool search and code mode, in the protocolRegistry ↗registry. modelcontextprotocol.

Hacker News · security

By rwilinski · mcp-explained.com

Read the full article at Hacker News · security

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.