---
title: "Comment by X1Y1 on \"An approved MCP tool definition is not a forever allow\" | Hashnode"
description: "Agree that the digest has to be checked on every tools/list refresh and not only at install. One thing I would add: a pin only helps if the fail-closed path is"
url: "https://hashnode.com/posts/an-approved-mcp-tool-definition-is-not-a-forever-allow/6abbcfe0f567119ea44d9904/comment/6ac5c4cb1d152f19c0b6bd87"
---

[Discussion](https://hashnode.com/posts/an-approved-mcp-tool-definition-is-not-a-forever-allow/6abbcfe0f567119ea44d9904)

Comment [An approved MCP tool definition is not a forever allow](https://hashnode.com/posts/an-approved-mcp-tool-definition-is-not-a-forever-allow/6abbcfe0f567119ea44d9904)

[X1Y1](https://hashnode.com/@x1y1)

[Oct 7](https://hashnode.com/posts/an-approved-mcp-tool-definition-is-not-a-forever-allow/6abbcfe0f567119ea44d9904/comment/6ac5c4cb1d152f19c0b6bd87)

Agree that the digest has to be checked on every tools/list refresh and not only at install. One thing I would add: a pin only helps if the fail-closed path is actually tested, so I would make the host log the old and new digest side by side when it refuses a call. Whoever gets paged then sees what changed without digging. Did Pillar say if Deadbugz changed only the descriptions, or the input schemas as well?