Meeting Transcript Analyzer Skill

Description

This skill processes meeting transcripts (from Zoom, Google Meet, Teams, Otter.ai, or plain text) to extract structured insights. It identifies action items, decisions made, questions raised, speaking time distribution, sentiment patterns, and key discussion topics. Outputs formatted meeting notes suitable for team distribution.

Instructions

1. Ingest: Accept transcript in common formats (VTT, SRT, TXT, JSON)
2. Parse: Identify speakers, timestamps, and utterance boundaries
3. Analyze: Extract action items, decisions, questions, and topic segments
4. Measure: Calculate speaking ratios, participation equity, and sentiment
5. Report: Generate structured meeting notes with all findings

Rules

• Preserve speaker attribution for all action items and decisions
• Distinguish between action items (someone commits to doing something) and suggestions (ideas mentioned but not committed to)
• Include timestamps for key moments so people can review the recording
• Calculate speaking time ratios to highlight participation imbalances
• Detect and flag unresolved questions (questions asked but not answered)
• Keep the summary factual -- do not add interpretations or opinions
• Handle multiple transcript formats gracefully
• Redact sensitive information if instructed (names, numbers, etc.)

Analysis Categories

Action Items Detection

Look for patterns like:

• "I'll [verb]..." / "I will [verb]..."
• "Can you [verb]..." / "Could you [verb]..."
• "Let's [verb]..." / "We should [verb]..."
• "[Name] will take care of..."
• "Action item: ..."
• "By [date/day]..."

Decision Detection

Look for patterns like:

• "We decided to..."
• "Let's go with..."
• "The decision is..."
• "We're going to..."
• "Agreed: ..."
• Consensus moments (multiple people agreeing)

Output Format

Copy
# Meeting Notes: [Meeting Title]
**Date:** [Date] | **Duration:** [Length] | **Participants:** [Count]

## TL;DR
[2-3 sentence summary of the meeting's purpose and outcome]

## Action Items
| # | Owner | Action | Due Date | Status |
|---|-------|--------|----------|--------|
| 1 | @alice | Update the API docs for v2 endpoints | March 28 | Pending |
| 2 | @bob | Set up staging environment for load testing | March 30 | Pending |
| 3 | @carol | Schedule design review with frontend team | Next sprint | Pending |

## Decisions Made
1. **Go with PostgreSQL over MongoDB** for the new service (timestamp: 12:34)
   - Rationale: Better suited for relational data, team has more experience
2. **Push launch date to April 15** (timestamp: 23:45)
   - Rationale: Need more time for security audit

## Unresolved Questions
- Who will handle on-call rotation for the new service? (raised by Bob at 18:20)
- What's the budget for additional infrastructure? (raised by Carol at 25:10)

## Key Discussion Topics
| Topic | Duration | Participants | Sentiment |
|-------|----------|-------------|----------|
| API v2 migration plan | 12 min | Alice, Bob, Dave | Positive |
| Database selection | 8 min | All | Mixed → Resolved |
| Launch timeline | 15 min | Alice, Carol | Concerned → Aligned |
| On-call rotation | 5 min | Bob | Unresolved |

## Participation Metrics
| Participant | Speaking Time | % of Total | Contributions |
|-------------|-------------|-----------|---------------|
| Alice | 14 min | 35% | 3 decisions, 1 action item |
| Bob | 10 min | 25% | 2 questions, 1 action item |
| Carol | 8 min | 20% | 1 decision, 1 action item |
| Dave | 6 min | 15% | 2 suggestions |
| Eve | 2 min | 5% | 1 agreement |

## Raw Notes (Chronological)
[Timestamped key points from the transcript]

Examples

"Analyze this Zoom transcript and extract action items"
"Generate meeting notes from this Google Meet recording transcript"
"What decisions were made in this meeting? Who owns what?"
"Show me speaking time distribution for this team meeting"
"Identify all unresolved questions from this standup transcript"