What it does
Meeting Summarizer parses a raw transcript — whether exported from Zoom, Google Meet, or Microsoft Teams — and produces a clean, structured Markdown document. It identifies speaker turns, locates decision statements, extracts action items with their assigned owners and due dates, and surfaces the key discussion points worth remembering.
The skill understands common transcript formatting conventions (speaker labels, timestamps, automated captions) and normalises them before analysis, so you get consistent output regardless of which platform generated the source file.
Example prompt
Summarize this meeting transcript: [paste transcript here]
Output format
The skill always produces output in the same section order so your team can find what they need instantly:
- ## Decisions — numbered list of final decisions reached during the meeting
- ## Action Items — table with columns: Task, Owner, Deadline
- ## Key Points — brief bullet list of notable discussion items that did not result in a decision or action
- ## Attendees — names detected from speaker labels
Supported transcript sources
- Zoom auto-generated transcripts (.vtt and plain text exports)
- Google Meet transcripts (Docs export)
- Microsoft Teams call transcripts (.docx and plain text)
- Otter.ai, Fireflies, and similar services (plain text format)
- Any transcript with "Speaker: text" or timestamped line format
Configuration options
You can adjust the skill's behaviour with optional flags appended to your prompt:
# Include raw speaker attributions for decisions Summarize this transcript --attribute-decisions # Output in a specific language Summarize this transcript --lang=Spanish # Limit summary to action items only Summarize this transcript --only=actions
Privacy note: Meeting Summarizer processes your transcript locally within Claude Code. No transcript content is sent to any third-party service. For sensitive meetings, this is the recommended approach over cloud-based summarisers.
Installation
# Install via npx npx skills add user/meeting-summarizer # Verify installation npx skills list | grep meeting
Frequently asked questions
Does it work with recordings instead of transcripts?
This skill works with text transcripts only. To use it with an audio or video recording, first run the recording through a transcription tool such as Whisper or your meeting platform's built-in transcription, then pass the resulting text to Meeting Summarizer.
How long can a transcript be?
The skill handles transcripts up to Claude's context window limit — typically a full 90-minute meeting falls well within this. For very long all-day sessions, split the transcript into logical segments (morning/afternoon) and summarize each separately, then combine.
Can it identify deadlines mentioned in passing?
Yes. The skill is trained to recognise relative date references ("by end of week", "before the sprint", "next Tuesday") as well as explicit calendar dates. It normalises these against the transcript date when available.