---
title: "Discussion on \"Watch Me Poison Your MCP\" | Hashnode"
description: "Discussion on \"Watch Me Poison Your MCP\". How MCP tool poisoning hijacks agent inference through description metadata, conversation-formatted JSON spoofs safety training."
url: "https://hashnode.com/posts/watch-me-poison-your-mcp/6995e7a48d33b96041da3239"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Watch Me Poison Your MCP](https://toxsec.hashnode.dev/watch-me-poison-your-mcp?utm_source=hashnode&utm_medium=feed)

How MCP tool poisoning hijacks agent inference through description metadata, conversation-formatted JSON spoofs safety training.

Feb 18 · 6 min read · toxsec.hashnode.dev  
Read the article

-   [#ai](https://hashnode.com/tag/ai)
-   [#infosec-cjbi6apo9015yaywu2micx2eo](https://hashnode.com/tag/infosec-cjbi6apo9015yaywu2micx2eo)
-   [#cybersecurity](https://hashnode.com/tag/cybersecurity)
-   [#genai](https://hashnode.com/tag/genai)
-   [#technology](https://hashnode.com/tag/technology)

## Written by

[ToxSec](https://hashnode.com/@ToxSec)

@ToxSec

  
[View profile](https://hashnode.com/@ToxSec)

Cybersecurity

M.S. Cybersecurity, CISSP. Ex-NSA, USMC.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.