granola-series-recaplisted

# granola-series-recap A single Granola meeting is a `type: event` memory. A *series* of related meetings is a **surface entity** — the `granola-series` kind. This skill clusters one-off meetings into series and turns each series into a richer retrieval target. ## Goal When the user asks "summarize my 1:1s with Alex" the kit should answer via the `[[Weekly Jane <> Alex 1:1]]` surface entity and its linked event memories — not by keyword-matching "Alex" across hundreds of memories. ## Read (before saving) 1. **Cluster meetings into series.** A series exists when: - Same title pattern across meetings (substring match) - Same attendee set - Regular cadence (weekly / bi-weekly / monthly) Heuristic to discover series from existing CAL/GRANOLA memories: ```python from collections import Counter titles = [m["title"] for m in calendar_or_granola_memories] normalized = [strip_dates(t) for t in titles] # remove "Apr 22" etc. recurring = [t for t, n in Counter(normalized).items() if n >= 3] ``` 2. **Does a granola-series surface entity exist for this cluster?** ``` entity_resolve "<series name>" ``` If no, create it under `entities/surfaces/granola-<series-slug>.md`. 3. **Check existing event memories** that should be re-linked to the newly-created series. ## Reflect Two memory outputs per series: - **Per-meeting** `type: event` memories (already exist for one-offs; this skill *re-links* them to the series surface) - **Per-serie