Tinder-style email triage that trains AI on your preferences. On activation call get_skill_context (MCP) first — one UI (index.html), demo is sample mail onl...
- Initial release of email-swipe: a Tinder-style email triage skill that trains AI on user preferences. - Supports swiping actions: left (don't keep), right (keep), and double-tap (important) to help build agent training packs. - Runs a single UI (`index.html` via `serve-ui.py`); demo mode uses sample mail on the same UI. - Requires MCP integration with `get_skill_context` for activation, mail access check, and guided agent intake process. - Offers multiple training paths: bootstrap (from scratch), calibrate (import rules), refine (target weak spots), and import-sorting (no swiping, learn from folders). - Emphasizes backend-first operation: prioritize inbox analysis and rule-building over UI-driven workflow.