Anki Add Notes

Anki Add Notes

Overview

This skill provides a procedural workflow for augmenting Anki note fields by taking existing note data, formatting it into a prompt template read from a file, and using your AI capabilities to generate content to populate a specific target field. It strictly targets notes where the target field is currently empty, preventing redundant work and data overwrites.

Parameters

When this skill is triggered, respect the following parameters (usually provided as flags or in the request):

• --deck: (Required) The Anki Deck Name to process.
• --target-field: (Required) The name of the field where the generated text will be stored (e.g., 'Notes', 'Mnemonic').
• --prompt-file: (Required) Path to a text file containing the prompt template. The template should use {FieldName} placeholders.
• --dry-run: (Optional) If enabled, identify and list the notes to be updated without actually processing or modifying them.
• --interactive / -i: (Optional) If enabled, process notes interactively, prompting the user to review and accept each generated response before updating.
• --limit: (Optional) Maximum number of notes to process in one run.
• --sort-field: (Optional) The field to use for sorting notes prior to processing. Defaults to Frequency.

Workflow

To augment Anki notes using these parameters, follow these steps exactly:

Step 1: Read the Prompt Template

Use your available file-reading tools to read the contents of the file specified in --prompt-file. Identify all required fields by finding placeholders in the format {FieldName} within the template.

Step 2: Fetch and Sort Notes

Query AnkiConnect locally at http://localhost:8765.

1. Use the findNotes action with the query deck:"<deck-name>".
2. Fetch the details of the returned notes using the notesInfo action.
3. Sort the resulting list of notes numerically by the value in the --sort-field (if the note lacks this field or it is empty, treat its value as 999999 for sorting purposes).

Step 3: Filter Notes

Iterate through the sorted notes to filter them:

1. Verify that the --target-field and all required fields (identified in Step 1) exist on the note type. If they do not exist, report an error and stop.
2. Keep only the notes where the --target-field is empty (or contains only whitespace).
3. If --limit is specified, limit the final list of filtered notes to this number.

Step 4: Handle Dry Run

If --dry-run is active, print a clear preview list of the filtered notes that would be updated (including their IDs and a preview of their relevant source fields), and then terminate the workflow without modifying any notes.

Step 5: Generate Content

For each filtered note:

1. Extract the actual field values from the note's fields.
2. Substitute the {FieldName} placeholders in the prompt template with these actual values to create the final prompt.
3. Generate the response directly using the filled prompt.
4. If --interactive (or -i) is enabled, present the generated text to the user for approval. Provide options: accept (y), reject (n), skip remaining (s), or quit (q). Respect their choice.
5. Convert the Markdown response to HTML. Ensure that all Markdown elements (like bolding, italics, lists, and newlines) are converted to their corresponding HTML tags (e.g., **text** becomes <b>text</b>, newlines become <br>). Specifically, replace any <strong> tags with <b> tags and </strong> with </b>.

Step 6: Apply Updates

Map the generated HTML content back to the original Note IDs. Update Anki by making requests to AnkiConnect:

1. Use the multi action to bundle multiple updateNoteFields and addTags requests into a single network call.
2. Set the --target-field with the new HTML content and add the tags anki_deck_augment and the current date (in YYYY-MM-DD format).

Guidelines

• Strictly Empty Targets: NEVER process notes that already have content in the --target-field. This is the primary filtering rule.
• Cost/Context Awareness: Processing large numbers of notes sequentially can consume large amounts of context. Always summarize the number of notes to be processed and pause to confirm if the count is large (>10 notes) before starting the generation loop, unless it's a dry run.