fluent-fsrs-referencelisted

# FSRS Scheduling Reference Fluent schedules reviews with **FSRS-6**, implemented in `.claude/hooks/fsrs.py` and driven by `.claude/hooks/update-db.py`. **Do not compute intervals by hand.** Unlike the old SM-2 formula, FSRS-6 uses 21 fitted weights, `stability`, and `fsrs_difficulty`; any hand calculation will diverge from the code. Skills submit a score/quality and let `update-db.py` do the scheduling. ## The pipeline ``` tutor score (0-10) → quality (0-5) quality = floor(score / 2) → rating (1-4) 1 if score<=4, 2 if <=6, 3 if <=8, else 4 → fsrs.schedule(...) → interval_days + due_date ``` You send `{ "item_id": "...", "quality": <0-5> }` (optionally `"score": <0-10>`) in `review_results[]`. `update-db.py` maps it to an FSRS rating and reschedules. ## Fields on a spaced-repetition item | Field | Meaning | |-------|---------| | `quality` / `last_quality` | 0-5 grade; feeds the mastery heuristic | | `repetitions` | consecutive-success counter; feeds mastery | | `mastery_level` | 0-5 stars, derived from repetitions + quality | | `stability` | FSRS memory stability (days) | | `fsrs_difficulty` | FSRS item difficulty (NOT the CEFR `difficulty` key) | | `interval_days` / `due_date` | computed by FSRS, do not set by hand | Items created before the FSRS migration may still carry a legacy SM-2 `easiness_factor`. Nothing reads it and new items no longer get one — ignore it. ## When to use Load when a skill must explain or reason about scheduling. To actually