← ClaudeAtlas

meeting-commitmentslisted

Track what the person promised in their meetings, suppress what is already handled, and hand back a short ranked list of what they still owe. Use when someone asks what they committed to, what they owe people, what they forgot, what is outstanding from their calls, "collect everything I was supposed to do", "what did I promise", "what am I behind on", "what's still open from my meetings", or asks for a recurring follow-up list. Also use to set that list up as a scheduled task, to mark commitments done, and on first run to configure where the tracker stores its data. Works against any meeting-notes source that exposes transcripts or summaries (Granola and Fireflies are used below as the worked examples). This tracks commitments across meetings over time; it does not write up an individual meeting.
alectivism/organization-ai-skills · ★ 0 · AI & Automation · score 75
Install: claude install-skill alectivism/organization-ai-skills
# Meeting Commitments Sweep recorded meetings for promises the person made, suppress the ones already handled, and return the ten that matter. Runs on a schedule without repeating itself. The split that makes this work: **you do the extraction, the ledger does the bookkeeping.** Judging whether "yeah I'll get that to you" is a real commitment needs a model. Remembering that it was already closed on Tuesday does not, and a model asked to remember it across runs will get it wrong. Never reconstruct what is outstanding by reading the previous output file. ## Any meeting-notes source works The examples below name Granola and Fireflies because they cover the common case of an AI notetaker. The mechanism only needs a source that can give you, per meeting: a date, the attendee or speaker list, and either a transcript or a summary with per-person action items. If your organization uses a different notetaker, a call-recording platform, or hand-typed meeting minutes in a document store, substitute its equivalent calls wherever Granola or Fireflies are named and keep the rest of the pipeline as written. ## Step 0: First run, or every run? Look for `config.json` in the person's store. If it is missing, this is a first run: go to **Setup** at the bottom of this file and do the interview. Do not guess a configuration and start sweeping. If it exists, read it and continue. The config names the store, the lookback window, the display cap, their categories, their excluded meetings, and