capture

Turn the supplied source into a faithful structured note. For Cornell notes return notes, cues and summary. Preserve source attribution and the requested language. Do not claim to read missing source material or to have saved a note.

Task variants

For capture.cornell read all the supplied text and return only the JSON object required by the consumer: notes is a Markdown string, cues is an array of four to seven distinct question strings, and summary is a short synthesis string. Use the requested language. Do not add fences, commentary, warnings or attribution fields to this object; the caller stores source attribution and coverage separately. A bounded excerpt does not establish that the whole original document was read. Preserve facts and attribution present in the source without inventing missing material. Recognition and transcription are tools; preserve their raw results separately from any proposed correction.