lorekeeper-protocollisted

# Lorekeeper Protocol Follow this protocol every session to keep your memory store accurate, healthy, and growing. --- ## Phase 1 — Session Start **Trigger**: Beginning of every session, before any substantive work. ### Steps 1. Identify the session topic (1–3 keywords). 2. Search for context: ``` lore_search({ query: "<topic>", min_score: 0.15, include_links: true }) ``` 3. Read all returned memories — decisions made, constraints, past patterns. 4. Provide feedback on every result immediately: ``` lore_update({ memory_feedback: [ { id: "<id>", useful: true }, { id: "<id>", useful: false } ] }) ``` Mark `useful: true` if relevant to this session. `useful: false` for noise. Skip only if zero results. 5. If fewer than 3 results, run a broader fallback search with related terms. **Do not skip Phase 1.** Working without context causes duplicate inserts and contradictory decisions. --- ## Phase 2 — Mid-Session (Topic Shift) **Trigger**: Conversation shifts to a new domain, subsystem, or question. 1. Detect the shift. 2. Run a fresh `lore_search` for the new topic — same pattern as Phase 1. 3. Provide feedback on the new result set. **Rule**: If you're reasoning about something you haven't searched for, search first. --- ## Phase 3 — Session End **Trigger**: End of every session. 1. **Insert new memories** for anything discovered: - Decisions (with rationale), bugs fixed (root cause), architecture insights - User corrections — strongest learn