Catalog v02 proved the healthy MCP schema pattern: explicit schemas, structuredContent, read-only annotations, fresh tool names, and clear non-mutating descriptions remove unsafe/unclear ChatGPT connector badges.
Catalog v02 proved the healthy MCP schema pattern: explicit schemas, structuredContent, read-only annotations, fresh tool names, and clear non-mutating descriptions remove unsafe/unclear ChatGPT connector badges.
afo-mcp-schema-hygiene · approved · 2026-05-30T03:38:43.759Z
#mcp#schema-hygiene#output-schema#read-only-tools#catalog-v02#chatgpt-connectors#tool-safety#structured-contentPreview
Prompt: And look at the screenshot from chat-gpt… healthy Response: The Catalog v02 screenshot confirms the schema-quality test worked. The tool badges now show READ instead of PUBLIC WRITE, OPEN WORLD, DESTRUCTIVE, UNCLEAR ARGUMENTS, or OUTPUT SCHEMA RECOMMENDED. That means ChatGPT understood these tools as read-only, non-destructive, structured, and safer to expose. The winning formula is: new MCP identity, fresh tool name