anki-study

name	anki-study
description	Teach the agent how to run tutor-style Anki study sessions.

Use the primary study workflows first and keep tutoring separate from collection mutation.

Prefer:

• anki_collection_status
• anki_study_start
• anki_study_card_details
• anki_study_reveal
• anki_study_grade
• anki_study_summary

Rules

1. Start with the narrowest deck or scope preset that matches the user goal, and compare due, new, learning, and review counts when the user needs help picking a deck.
2. Use anki_study_card_details as the default current-card read for front-side tutoring, and use anki_study_reveal when the user asks to reveal the answer or back side.
3. Present study cards like Anki by default: focus on the prompt, question, and front-side clues first, and do not volunteer the answer or back-side content until after you call anki_study_reveal; include media in the same response when it helps the learner without spoiling the answer.
4. Present grading choices as a numbered list in this order when you guide the user: 1. Again 2. Hard 3. Good 4. Easy.
5. Treat study-session state as the tutoring source of truth, and only call anki_study_grade after the current card has been revealed.
6. Prefer curated study_view output over raw_fields, and only fall back to raw fields when the curated answer is missing or the user asks for full detail.
7. When current_card.view is present, rely on the returned canvas metadata to show the card in Control UI, use current_card.tutoring_summary for reasoning, and use current_card.study_media for native Discord or Telegram media delivery; if media resolution fails, name the structured error code instead of inventing a generic failure.
8. Explain misses in study terms and patterns instead of dumping raw database fields.

Anti-Patterns

• Do not default to low-level tools when the user asked to study.
• Do not mutate notes or decks unless explicitly asked.
• Do not skip anki_study_reveal when the user asks for the answer or when grading requires a revealed card.