updating-thoughts

Edits existing Open Brain thoughts via the update_thought MCP tool with metadata patching and content rewriting.

4.5k|872|Updated Mar 11, 2026
One-click install
npx skills add https://github.com/NateBJones-Projects/OB1 --skill updating-thoughts
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: updating-thoughts
Source: https://github.com/NateBJones-Projects/OB1/tree/main/skills/updating-thoughts
Command: npx skills add https://github.com/NateBJones-Projects/OB1 --skill updating-thoughts

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

The core Open Brain MCP server can capture and read thoughts but cannot change them. This Skill teaches an AI client how to correctly edit an existing thought: resolving its UUID first, then choosing between a cheap metadata tag update and a full content rewrite that re-embeds the text for semantic search.

Core Features & Use Cases

  • Metadata patching: Add or change tags, status flags, and cross-references with metadata_patch, which shallow-merges keys without re-embedding.
  • Content rewriting: Replace a thought's text with the content field, which re-embeds it so semantic search stays accurate.
  • Optimistic concurrency: Guard read-modify-write cycles with if_unchanged_since, which rejects stale writes with a STALE_READ error.
  • Use Case: Ask your AI client to "mark my thought about the Q3 roadmap as superseded" — it searches for the thought, resolves its UUID, applies a metadata-only patch, and confirms the new updated_at without wasting an embedding call.

Quick Start

Ask your connected AI client to add a status tag to your thought about a specific topic, and it will search for the thought, resolve its UUID, and apply a metadata patch via the update_thought tool.

Frequently Asked Questions about updating-thoughts

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I update an existing thought in Open Brain?

First resolve the thought's UUID using search_thoughts or list_thoughts, then call update_thought with that id. Pass metadata_patch for tag changes or content for text rewrites, and the tool returns a confirmation with the new updated_at timestamp.

What is the difference between metadata_patch and content in update_thought?

metadata_patch shallow-merges keys into the thought's metadata without re-embedding, making it cheap for tagging. content replaces the thought's text and triggers re-embedding so semantic search stays accurate.

Does metadata_patch replace the entire metadata object?

No, metadata_patch is a merge, not a replace. Passing {"status":"done"} adds or overwrites only the status key and leaves all other metadata keys intact.

Why did my thought update return a STALE_READ error?

STALE_READ means another writer changed the row after the updated_at timestamp you passed in if_unchanged_since. Re-fetch the thought to get the current state and retry the write with the fresh timestamp.

Which AI clients work with the update_thought MCP connector?

Any client that supports reusable skills, rules, or custom instructions and can connect to the open-brain-update-thought MCP server works, including Claude Code, Codex, and Grok. Authentication uses the same MCP_ACCESS_KEY as the core Open Brain connector.

When should I update a thought instead of deleting it?

Prefer updating or tagging a thought as superseded when it is outdated or wrong, since deletion removes context other thoughts may reference. Deletion is covered by the separate deleting-thoughts skill.