community-calls

Turn a call transcript into the committable call markdown — cleaned transcript + chapters + AI summary — without rendering any video. Content-aware: it inspects calls/<id>.md and runs only the stages not yet done (or the one you ask for). This is the "update the markdown" half of the call pipeline; use it directly when you want to process or fix a call's text without making a video.

Render the full call video (MP4) from a completed call markdown, and produce the paste-ready YouTube handoff block (download URL, title, description, chapters). This is the "make the video" half of the call pipeline; run it when the markdown is already done and you want the video. Renders the full final by default (no draft unless asked). Needs the distributable audio mp3 to exist.

Orchestrate the full call pipeline end to end — from a transcript to a published call page (cleaned transcript, chapters, AI summary) and a rendered video. Inspects the call markdown to see what's already done, proposes the remaining steps, and composes the per-stage skills, confirming before the render. Use for the whole flow; for a single part, run that step's skill directly.

Fix speaker attribution in a call transcript when one Zoom participant is actually two+ people (shared mic / same room). Local, CPU-only voice attribution that relabels the transcript and splits cues at speaker changes. On-demand — most calls do NOT need this; run it only when a Zoom label conflates multiple voices.

Draft an agenda for the next upcoming community call based on recent calls, GitHub activity, and ecosystem news

Schedule a new community call by creating a skeleton markdown file with date, time, and placeholder agenda