gbrain-think

Synthesize cited answers from gbrain knowledge pages with conflict and gap analysis.

1|Updated Jun 21, 2026
One-click install
npx skills add https://github.com/tapway/shogun-os --skill gbrain-think-tapway
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gbrain-think
Source: https://github.com/tapway/shogun-os/tree/main/skills/gbrain/think
Command: npx skills add https://github.com/tapway/shogun-os --skill gbrain-think-tapway

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Raw search returns isolated pages, leaving you to manually reconcile conflicting information and spot missing knowledge across your gbrain knowledge base. This Skill runs multi-hop synthesis to produce a single cited answer, flagging where sources disagree and what the brain does not yet know. ## Core Features & Use Cases - Multi-hop synthesis: Searches the brain, reads top results, and composes an answer with inline citations to source pages. - Conflict and gap detection: Reports when sources disagree (e.g., different dates for the same event) and lists knowledge gaps for follow-up research. - Temporal and entity controls: Supports since/until time windows, entity anchoring, and multi-pass rounds for deeper research. - Use Case: Ask "What do we know about Kossan's expansion plans?" and receive a synthesized briefing with citations, a note that March and June sources conflict on the opening date, and gaps like missing staffing plans — then chain a web search and write findings back to the brain. ## Quick Start Ask the agent to use gbrain think to answer "What do we know about X?" with citations, conflicts, and gaps.

Frequently Asked Questions about gbrain-think

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

FAQPage Schema
How do I get a synthesized answer from my gbrain knowledge base?

Call the mcp_gbrain_think MCP tool with your question, or run gbrain think from the CLI as a fallback. It searches the brain, reads top results, and returns a synthesized answer with inline citations, evidence pages, conflicts, and gaps.

When should I use gbrain think instead of gbrain search?

Use think for complex questions needing synthesis across multiple pages, such as briefings or status updates. Use mcp_gbrain_search for simple fact lookups or questions answerable from a single page, since think is heavier than raw search.

Does gbrain think support filtering by date or entity?

Yes. The MCP tool accepts since and until parameters (YYYY-MM-DD) to bound a temporal window, and an anchor parameter to pull the entity subgraph around a specific page slug. A rounds parameter enables multi-pass research.

What are the prerequisites for running gbrain think?

You need the gbrain MCP server connected with mcp_gbrain_think available, and the brain indexed with embeddings because think uses hybrid search internally. For CLI fallback, the gbrain think command must be working.

How do I fix knowledge gaps that gbrain think reports?

Chain the workflow: run think to get the gap analysis, search the web or other sources for the missing information, then write findings back to the brain with mcp_gbrain_put_page so future queries include the new knowledge.