A cleaned meeting has a transcript that reads well. Speaker names follow the full-name-first convention (full name on first occurrence, first name thereafter), filler words are calibrated to the chosen cleanup level, and formatting is consistent. The transcript is a reliable record of what was said.
Cleaning is editorial work, not synthesis. The goal is a transcript that someone can read and follow without guessing who said what or wading through verbal tics. The content structure still mirrors the conversation’s chronological flow — topical organization and insight extraction happen at the next stage.
Cleaned → Summarized: When the transcript has been synthesized into structured meeting notes — organized by topic rather than chronology, with action items extracted, key takeaways identified, and an inverted pyramid structure (summary above the divider, detail below). The /meeting-notes skill handles this step.
Processing track design from [[Linear Processing Stages for Meetings]]↑. Cleanup conventions from the /transcript-cleanup skill, which defines light/medium/heavy cleanup levels. Track placement within the three-track status system from [[Status Lifecycle Tracks]]↑.