What problem does it solve? Agents often waste API credits and time by searching the web, Slack, or email for information that already exists in the internal gbrain knowledge base of ~16,000 pages. This Skill enforces a mandatory lookup order so internal knowledge is always checked first. ## Core Features & Use Cases - Mandatory Lookup Chain: Defines a strict escalation order from keyword search (mcp_gbrain_search) through hybrid semantic search (mcp_gbrain_query) to multi-hop synthesis (mcp_gbrain_think) before any external API is allowed. - Quality Rules: Sets score thresholds (use results above 0.3), requires source slug citations, and mandates flagging knowledge gaps instead of hallucinating. - Sub-agent Propagation: Provides a standard instruction line to embed in delegated task prompts so sub-agents follow the same brain-first protocol. - Use Case: When asked "Tell me about Eddie Goh", the agent searches gbrain first, retrieves the person page with citations, and only reaches for web search if the brain returns nothing useful. ## Quick Start Ask the agent to look up any person, company, or fact and it will query gbrain before touching any external API.
