Anki Skill

Anki Flashcard Skill

Converts any source material — YouTube video, PDF, article, or pasted text — into Anki flashcards, with an optional self-quiz step to surface knowledge gaps first.

Trigger phrases

• "make Anki cards from…"
• "flashcard this…"
• "add to Anki…"
• "quiz me then Anki…"
• /anki <source>

Inputs accepted

Optionally append:

• --deck "DeckName" to choose the destination Anki deck (default: Claude::Imported)
• --quiz to enable the gap-finding quiz step before card generation
• --tags tag1 tag2 to attach Anki tags to every card

Workflow

Step 1 — Ingest the source

YouTube URL

python get_transcript.py "<url>" --format text

Read the output. It will be timestamped lines like:

[1:23](https://youtu.be/abc?t=83) The creator explains ...

Use the full transcript as the knowledge base.

PDF or text file Read the file directly with the Read tool.

Web article Fetch the URL with WebFetch.

Pasted text Use as-is.

Step 2 — (Optional) Gap-finding quiz

Only run this step if the user passed --quiz or asked to be quizzed first.

1. Identify the 8–12 most important concepts in the source.
2. Quiz the user on each one — ask a single question and wait for their answer before moving on.
3. Mark each answer as: solid / shaky / missing.
4. Only generate cards for shaky and missing concepts.
5. After the quiz summarise: "You knew X/Y topics solidly. Generating cards for the gaps."

If --quiz was NOT requested, skip straight to Step 3 and generate cards for all key concepts.

Step 3 — Generate flashcards

Produce a JSON array of card objects. Rules:

• Front: one atomic question — no compound questions.
• Back: concise answer (1–3 sentences max). Include a timestamp link if the source was YouTube so the user can jump straight to the relevant moment.
• Aim for 10–25 cards unless the source is very short or very long.
• Prefer "why/how" questions over pure definition recall.
• For sequential content (move sequences, algorithms, steps) use cloze-style fronts: "Step 3 of the euro-step is: ___"

Example output:

[
  {
    "front": "What is the key footwork principle behind the euro-step?",
    "back": "The second step plants in a different direction than the first, exploiting the gather rule. <a href='https://youtu.be/abc?t=142'>1:42</a>",
    "tags": ["basketball", "footwork"]
  }
]

Write this JSON to a temp file: /tmp/anki_cards.json

Step 4 — Preview

Show the user a formatted preview table:

Ask: "Add all N cards to Anki, edit any, or cancel?"

• Add all → proceed to Step 5.
• Edit → ask which card numbers to change, apply edits, re-preview.
• Cancel → stop, leave /tmp/anki_cards.json in place so they can inspect it.

Step 5 — Write to Anki

python anki_connect.py /tmp/anki_cards.json --deck "<deck>" --model Basic

Report the result: how many cards were added, any duplicates skipped.

If AnkiConnect is not running (connection refused), print the setup instructions below and offer to save the cards as a plain .txt import file instead.

AnkiConnect setup (first-time)

1. Open Anki desktop.
2. Tools → Add-ons → Get Add-ons → code 2055492159 → restart Anki.
3. AnkiConnect now runs on http://localhost:8765 whenever Anki is open.
4. Re-run this skill.

Helper scripts in this folder

Fallback: no AnkiConnect

If the user cannot use AnkiConnect, write the cards as a tab-separated .txt file:

front<TAB>back<TAB>tags

This format imports directly via Anki → File → Import.