comm-zoom-debrief

COMM-ZOOM-DEBRIEF — Live Zoom Meeting Debrief

Requires: Zoom MCP connector. All data is pulled live from Zoom — no transcript pasting needed.

---

Overview

This skill finds any meeting in the user's Zoom account, pulls the full AI summary, transcript, and auto-generated next steps, then structures everything into a professional debrief: decisions made, action items with owners and deadlines, follow-ups needed, and key insights. The structured debrief is saved to Cloud Brain and optionally pushed back into Zoom as a Zoom Doc the whole team can reference.

This replaces the manual workflow of copying transcripts out of Zoom and pasting them into a separate tool.

---

When This Skill Applies

• User says "debrief my last meeting"
• User says "process the call with [person/company name]"
• User says "what happened in the [topic] meeting?"
• User says "extract action items from my meeting"
• User says "pull up the recording from [date]"
• User says "post-meeting summary" or "meeting recap"
• User says "what did we decide in [meeting topic]?"
• User wants structured output from any Zoom call, whether it happened today or weeks ago

---

Pre-Flight — User Preferences

Before locating the meeting, load any saved communication preferences from Cloud Brain.

Search Cloud Brain using mcp__cloud-brain__search_notes with query: "zoom debrief preferences".

If preferences ARE found: Confirm in one line — "Using your saved preferences — [name], output style: [style]."

If NO preferences found: Ask these two quick questions before proceeding:

1. What is your name (so I can identify your action items vs. others')?
2. For action items assigned to you, would you like a default deadline added if none was stated? (e.g., "3 business days") — Yes/No

Save to Cloud Brain:

• Tool: mcp__cloud-brain__write_note
• Path: brain/preferences/zoom-debrief-preferences.md

---

Step 1: Identify the Meeting

Determine which meeting to process from the user's request.

Ask for user's current timezone before searching — this is required by the Zoom search API. If the user has preferences saved that include timezone, use that. Otherwise ask: "What timezone are you in? (e.g., America/Phoenix, America/New_York)"

Parse the user's request for meeting identifiers:

• Meeting topic or keywords ("the onboarding call", "the pricing discussion")
• Person's name ("my call with Sarah")
• Date or relative time ("yesterday", "last Tuesday", "the meeting on June 10")
• Meeting ID (if provided directly)

Search for the meeting using mcp__zoom__search_meetings:

• Set q to the topic/keywords or person's name
• Set from/to based on the date context (if "last week," set the range accordingly)
• If "most recent" or "last meeting," set from to 7 days ago

If multiple meetings match: Display a brief list and ask the user to confirm which one:

I found [N] meetings that might match:
1. [Topic] — [Date] — [Duration]
2. [Topic] — [Date] — [Duration]
3. [Topic] — [Date] — [Duration]

Which meeting would you like to debrief?

If no meeting matches: Ask the user to refine — "I couldn't find a match. Can you give me more detail about the meeting topic or date?"

---

Step 2: Pull Meeting Assets

Using the meeting_uuid from the search result, call mcp__zoom__get_meeting_assets.

This returns:

• meeting_summary — Zoom AI Companion's auto-generated summary (quick recap, full text, next steps)
• my_notes — Host/participant notes taken during the meeting
• recording — Cloud recording URL and processing status
• participants — Attendee list
• agenda_doc — Agenda document if one was attached

If recording is still processing (processing: true), note: "The recording is still being processed by Zoom. The AI summary and notes are available — I'll use those. If you want the full transcript, try again in a few minutes."

---

Step 3: Pull Recording Transcript and AI Next Steps

If recording is available, call mcp__zoom__get_recording_resource with types: "transcript,summary,nextStep" to get:

• Full timestamped transcript
• Zoom's auto-generated topic summaries
• Zoom's auto-generated next steps

Use both the AI summary AND the transcript. The summary is fast; the transcript catches specifics that the summary may gloss over.

If transcript is unavailable (recording not enabled, still processing, or no recording for this meeting): Proceed with AI summary and notes only. Flag: "No transcript was available — analysis is based on Zoom's AI summary and meeting notes."

---

Step 4: Extract Structured Output

Analyze all available data (AI summary, Zoom next steps, transcript, notes) and extract:

Decisions Made

• What was decided (the outcome, not the discussion)
• Who made or agreed to the decision
• Context or rationale if stated

Action Items

For each action item, capture:

Field	What to Extract
Action	What needs to be done (specific, clear)
Owner	Who is responsible (name or role)
Deadline	Extract if mentioned; apply default if user has one saved; otherwise "TBD"
Priority	High / Medium / Low based on urgency signals in conversation
Source	Brief quote or context from transcript where this was stated

Highlight action items assigned to the user (by name match against saved preferences). These appear first in the output under "Your Action Items."

Follow-Ups

• Items that need further discussion
• Questions raised but not resolved
• Topics deferred to a future meeting

Key Insights

• Strategic points worth remembering
• Notable commitments or concerns raised
• Any relationship-relevant context (what the client cares about, objections raised)

Content to EXCLUDE:

• Technical meeting issues ("you're on mute", "can you share your screen")
• Side conversations unrelated to the meeting topic
• Incomplete or inaudible/garbled transcript fragments

---

Step 5: Generate the Debrief Document

Format the structured debrief:

# Meeting Debrief: [Meeting Topic]
**Date:** YYYY-MM-DD  
**Attendees:** [Name 1], [Name 2], [Name 3]  
**Duration:** [X minutes]  
**Source:** Zoom recording — [meeting_uuid]

---

## Summary
[2–4 sentence executive summary of what the meeting was about and what was accomplished]

---

## Your Action Items
| # | Action | Deadline | Priority |
|---|--------|----------|----------|
| 1 | [Task — for the user specifically] | [Date/TBD] | High |

## All Action Items
| # | Action | Owner | Deadline | Priority |
|---|--------|-------|----------|----------|
| 1 | [Task] | [Name] | [Date/TBD] | [Level] |
| 2 | [Task] | [Name] | [Date/TBD] | [Level] |

---

## Decisions Made
| # | Decision | Made By |
|---|----------|---------|
| 1 | [What was decided] | [Person] |

---

## Follow-Ups
- [ ] [Item] — Owner: [Name] — By: [Date/Next meeting]
- [ ] [Question to resolve] — Owner: [Name]

---

## Key Insights
- [Strategic observation or important context]
- [Notable commitment, concern, or relationship insight]

---

*Processed by COMM-zoom-debrief. Source: Zoom meeting [meeting_uuid] on [date].*

---

Step 6: Save to Cloud Brain

Save the structured debrief to Cloud Brain:

• Tool: mcp__cloud-brain__write_note
• Path: brain/meetings/[slug-from-topic]-[YYYY-MM-DD].md
• Include full structured debrief content

Also check if there is an active project in the brain related to this meeting's topic:

• Search Cloud Brain: mcp__cloud-brain__search_notes with the meeting topic keywords
• If a project note is found, append a meeting reference:

  ## Meeting Update [YYYY-MM-DD]
  - [1–2 sentence summary of relevant outcomes]
  - Action items: [N items assigned]
  

---

Step 7: Create Zoom Doc (Optional)

After saving to Brain, ask: "Want me to also save this debrief as a Zoom Doc so your team can see it?"

If yes, call mcp__zoom__create_new_file_with_markdown:

• file_name: "Meeting Debrief — [Topic] — [YYYY-MM-DD]"
• content: the full structured debrief in markdown

Provide the returned file_link to the user.

---

Step 8: Confirm Results

Report to the user:

Debrief complete: [Meeting Topic] — [Date]
Attendees: [N] people

Extracted:
✅ [N] decisions
✅ [N] action items ([X] assigned to you)
✅ [N] follow-ups
✅ [N] key insights

Saved to: brain/meetings/[slug]-[date].md
[Zoom Doc link if created]

Your next actions:
1. [User's #1 action item] — [Deadline]
2. [User's #2 action item] — [Deadline]

---

Error Handling

• Meeting not found: Ask for more details (topic, date, attendee name). Offer to show a list of recent meetings.
• Recording not available / Zoom AI not enabled: Proceed with whatever data is available (summary, notes). Flag what's missing.
• No decisions or action items found: Report: "This meeting appears to have been informational — no clear decisions or action items were identified. Key discussion points are captured under Key Insights."
• Transcript still processing: Use AI summary and notes. Tell user to run the debrief again in 10–15 minutes for the full transcript.
• Permission denied on meeting assets: "I don't have permission to access this meeting's assets. You may need to be the host or have been granted access by the host."