spaced-repetition

Run the scheduler script with the --today flag: python scripts/srs_scheduler.py --today. It prints all items whose next_review date is today or earlier, sorted by priority, due date, and repetition count.

Scores of 3 or higher increase the interval (1 day, then 6 days, then interval times the easiness factor). Scores below 3 reset repetitions to zero and the interval to 1 day. The easiness factor adjusts between roughly 1.3 and 2.5 based on score.

0 means complete forgetting, 1-2 means wrong answer requiring relearning, 3 means correct but difficult, 4 means correct, and 5 means very easy. Lower scores also raise the item's review priority to high.

Yes, run the script with --sync-weak-points. It reads weak_points from learning.json, adds new ones as high-priority items, promotes existing items to high priority, and skips duplicates using normalized text comparison.

Review items persist in .agent/state/srs_log.json, and each save also syncs an SRS snapshot and last session date into learning.json. Both are plain JSON files you can inspect or back up.

Not yet in code. The SKILL.md documents a mapping of grading outcomes to initial due dates (D+1 to D+7), but the script currently fixes the initial interval at 1 day, so custom initial dues must be handled manually.