Skip to content

v0.41.1

Choose a tag to compare

View all tags
@helpcodeai helpcodeai released this 19 Oct 20:20
v0.41.1
af1319b
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Skills now reach Claude on claude.ai

claude.ai does not show an MCP server's instructions to the model, and it loads tool details only when it is about to use a tool. Because of this, workspace skills that were only delivered in the instructions never reached Claude there.

  • Skills in tool descriptions. Each tool now names the workspace skills that cover it. When there are only a few, each one is listed with when to use it, its rule, and a pointer to skill_get if it has a full procedure. When there are more, the tool shows how many there are and points to skills_list. Every client reads these descriptions, so the skills apply wherever the tool is used.
  • Clearer skill learning. For members who contribute to learning, _intent now asks for one short sentence about what the user wants to do, with no personal data. It also explains why it is asked and that the user can turn learning off in their profile. Before this, models often left the field empty.
  • Memory items saved from an AI client no longer show the sign-in method as the client name.

After upgrading, refresh each AnythingMCP connector in claude.ai so new chats get the current tools: Customize → Connectors → the connector → ⋮ → Refresh tools list.

No database migrations. Full changes: #1166, #1164.

Assets 2
Loading