---
name: meeting-summary-template
description: "Turn a raw transcript into a decision-focused recap: outcomes, owners, deadlines, open threads."
title: Meeting Summary Template
category: productivity-ops
difficulty: beginner
license: MIT
author: admin
source_url: "https://github.com/anthropics/anthropic-cookbook"
icon: 📝
input: text
output: markdown
phase: transform
domain: ops
tags: meeting-minutes,transcript-processing,decision-extraction,action-items,prompt-schema,structured-output,noise-filtering,owner-assignment,deadline-tracking,markdown-formatting,whisper-json,follow-up-optimization
best_for:
  - Post-meeting recaps for distributed teams
  - Converting Whisper/Otter transcripts to actionable minutes
  - Distilling decisions and owners from unstructured discussion logs
  - Streamlining follow-up workflows and accountability tracking
---

## Description

A prompt + schema skill that converts meeting transcripts into structured minutes. Enforces four mandatory sections: Decisions Made, Action Items (with owner + due date), Open Questions, and Out-of-Scope. Optimized to strip filler and speculative talk.

## Why it works

Unstructured meeting summaries are useless a week later because they mirror the transcript's chaos. Decisions aren't what was discussed — they're what everyone agreed to. Action items aren't tasks raised — they're tasks *claimed* by a named owner. Forcing the template to reject anything without an owner or a decision verb filters 80% of the noise and leaves the 20% that drives follow-up.

## How it works

1) Ingest transcript (Whisper JSON with speaker labels is the sweet spot). 2) First pass: flag every sentence containing decision language ('we'll', 'agreed', 'decided', 'blocked on'). 3) Second pass: for each flagged sentence, resolve the speaker, the object, and a due-date if mentioned. 4) Items without a clear owner get routed to 'Open Questions' instead of silently dropped. 5) Final output is markdown: ## Decisions, ## Action Items (table: owner, task, due), ## Open Questions, ## Out-of-Scope. Re-queries the LLM once to tighten verb tense + remove 'we should think about' hedge phrases.