Search the agent-tome knowledge base for prior research. TRIGGER as a first step when asked a knowledge question, research question, or 'what is X' question — check the tome BEFORE answering from memory or doing web research. Also trigger when the user asks about prior findings or references. TRIGGER when preparing for meetings, reviews, or briefings where domain context would help. Even when the user provides specific links or documents to review, check the tome first for related domain context that could accelerate synthesis.
You are checking the agent-tome knowledge base for existing knowledge before doing fresh research. The knowledge base is append-only and contains accumulated research findings, reports, and references.
Keywords are the only discovery mechanism — there is no full-text search. Start with a keyword prefix search to understand the vocabulary:
agent-tome keywords <prefix>
Try a few prefixes related to your topic. Keywords are singularised and downcased (e.g., "threads" is stored as "thread").
agent-tome search <keyword1> <keyword2> ...
Default is --match any (OR). Use --match all to narrow results when you get too many hits.
Results are ranked by number of matching keywords. Each result shows global_id, description, keywords, and matching_keyword_count.
agent-tome fetch <global_id>
Returns full content: all entries (chronologically ordered), sources (web and file), and consolidation history. Read the entries to understand what's already known.
agent-tome related <global_id>
Finds articles connected through shared keywords, explicit references, and consolidation links. Useful for discovering adjacent knowledge you didn't search for directly.
agent-tome source-search <url-or-path>
If you're about to fetch a URL, check whether it's already been consulted. URLs starting with http:// or https:// are matched as web sources; everything else as file paths.
After looking up the knowledge base, report to the user: