Meeting Summary Generator

Extract action items, decisions, and key points from meeting transcripts. Automatically routes items to tasks.md or GitHub Issues.

Usage

/summarize_meeting [transcript_file]
/summarize_meeting .research/meetings/transcripts/2024-12-02-lab-meeting.md

When to Use

• After running /transcribe
• On any meeting transcript or notes
• To process handwritten meeting notes (type them first)

Prerequisites

• Transcript exists in .research/meetings/transcripts/ folder
• Transcript is in markdown format

Execution Steps

1. Load Transcript

Read the meeting transcript and project context:

• .research/meetings/transcripts/[filename].md - The transcript
• .research/project_telos.md - Project aims (for context)
• tasks.md - Current tasks (avoid duplicates)

2. Extract Key Information

Analyze transcript for:

• Decisions made
• Action items (who, what, when)
• Questions raised
• Key insights or ideas
• Follow-up needed

3. Generate Meeting Summary

Append summary to transcript or create separate file:

---

# Meeting Summary

## Key Decisions
<!-- Decisions that were made during the meeting -->
1. [Decision 1] - [Brief context]
2. [Decision 2] - [Brief context]

## Action Items

### Tasks (< 2 hours, single implementation)
<!-- These will be added to tasks.md -->

| Item | Owner | Due | Priority |
|------|-------|-----|----------|
| [Task description] | [Name/@you] | [Date/ASAP] | [High/Med/Low] |
| [Task description] | [Name/@you] | [Date] | [Priority] |

### Issues (> 2 hours, needs tracking)
<!-- These will become GitHub Issues -->

1. **[Issue title]**
   - Description: [What needs to be done]
   - Why: [Why this is needed]
   - Complexity: [Estimate]
   - Labels: [Suggested labels]

2. **[Issue title]**
   - Description: [Details]
   - Why: [Rationale]

## Key Insights
<!-- Important points or ideas worth remembering -->
- [Insight 1]
- [Insight 2]

## Open Questions
<!-- Questions that weren't resolved -->
- [ ] [Question 1] - Needs: [Who/what to resolve]
- [ ] [Question 2] - Needs: [Who/what to resolve]

## Follow-up Needed
<!-- Things to discuss or check on later -->
- [Follow-up item]

## Next Meeting
<!-- If discussed -->
- **Date**: [If scheduled]
- **Agenda items**: [If mentioned]

---

*Summary generated: [Timestamp]*

4. Task vs Issue Classification

Apply this heuristic:

Criteria	→ Task (tasks.md)	→ Issue (GitHub)
Estimated time	< 2 hours	> 2 hours
Scope	Single action	Multiple steps
Branching	Not needed	Needs own branch
Comparison	No	Comparing alternatives
Documentation	Not needed	Should be tracked
Project direction	Doesn't change	May change direction

When uncertain, ask:

I found this action item: "[Item description]"

This could be:
A) A quick task (< 2 hours, add to tasks.md)
B) A larger issue (needs GitHub Issue for tracking)

Which fits better? (Or provide more context)

5. Update tasks.md

Add new tasks with meeting reference:

## From Meeting: 2024-12-02-lab-meeting

- [ ] [Task 1] (Due: [date])
- [ ] [Task 2] (Due: [date])
- [ ] [Task 3]

6. Create GitHub Issues

For items classified as issues, offer to create:

I identified 2 items that should be GitHub Issues:

1. "Compare SVM vs Random Forest for classification"
   - Would require testing both approaches
   - Results should be documented for paper
   
2. "Implement alternative normalization method"
   - Needs research into options
   - May change downstream pipeline

Create these as GitHub Issues? (Y/n)

If yes, create issues with:

• Clear title
• Description from meeting context
• Labels (if determinable)
• Reference to meeting transcript

7. Post-Summary Actions

Meeting summarized!

Summary added to: .research/meetings/transcripts/2024-12-02-lab-meeting.md
Tasks added: 3 new items in tasks.md
Issues to create: 2 (awaiting confirmation)

Next steps:
A) Create the GitHub Issues
B) Review and prioritize new tasks
C) Update project_telos.md with decisions made
D) Continue with other work

What would you like to do?

Example Output

# Meeting Summary

## Key Decisions
1. Use random forest as primary classifier (SVM as comparison)
2. Deadline for analysis: end of month
3. Weekly check-ins moving to Tuesdays

## Action Items

### Tasks
| Item | Owner | Due | Priority |
|------|-------|-----|----------|
| Fix axis labels on Figure 2 | @you | Dec 4 | Low |
| Send PI the draft methods section | @you | Dec 3 | High |
| Update README with new instructions | @you | Dec 5 | Med |

### Issues
1. **Compare SVM vs Random Forest performance**
   - Description: Run both classifiers with same CV scheme, compare metrics
   - Why: Reviewer may ask about method choice
   - Complexity: ~4-6 hours
   - Labels: analysis, methodology

## Key Insights
- PI suggested looking at recent paper by Smith et al. on normalization
- Feature importance might be more interesting than just accuracy

## Open Questions
- [ ] Which normalization method should we use? - Needs: literature review
- [ ] Include supplementary figures in main text? - Needs: check journal guidelines

Related Skills

• transcribe - Generate transcript from audio
• weekly-review - See tasks in context of weekly work
• plan-week - Incorporate new tasks into weekly plan
• next - Get next suggestion

Notes

• Review summaries for accuracy - AI may misinterpret discussion
• Action items should have clear owners
• Link issues back to meeting for context
• Consider who said what for attribution