classify-notelisted

# classify-note The classifier. Two decisions in one pass: which `kind` (focus, doing, habit, knowledge, contact/person, contact/organization), and which bundle the material belongs to. ## Directives 1. Bundle is the default. Material that shares a theme (a hobby, a project, a recurring topic) lives in one bundle under `<kind>/<theme-slug>/`. Singles are the exception, not the default. 2. Existing bundle wins. If `<kind>/<theme-slug>/` already exists and the new material fits the theme, file there. Do not create a sibling bundle for the same theme. 3. Knowledge is the default kind for ambiguous cases. Mis-classifying as knowledge is recoverable. Mis-classifying as focus pollutes attention. 4. Time semantics matter. Past-dated material is knowledge or archive, not focus. 5. Pattern-match intent, not literal strings. A sentence in the user's writing language that expresses a near-term commitment is focus material, regardless of which exact words appear. 6. Use the index, not grep. Bundle match and theme derivation read `zanmai/memory/patterns.json` via `zanmai.py index find`. No `grep -r` over the vault, no body read across many files. The index is regenerated on demand via `zanmai.py index rebuild` plus `zanmai.py index patterns`. ## When to use - Inside `import-bundle` Step 2 (kind plus bundle decision). - When the user files standalone material via Steve. - When deciding where new material attaches to existing bundles. ## When not to use - The user explicitly said bot