Meeting Facilitation skill for Claude Code
Meetings that end in decisions and named owners. Claude first checks whether a meeting is needed at all, then writes an agenda of timeboxed questions with an outcome sentence, a named decider and a decision method. During the meeting it applies concrete moves for quiet voices, tangents and circular debate, and afterwards it sends a record that leads with decisions, trade-offs, dissent and actions owned by one person each.
- name
meeting-facilitation - size180 lines · ≈2,000 tokens
- files1
- licenseMIT
- updated
What Claude sees before loading it
This description is the only part of the skill that sits in context all the time. Claude reads it to decide when to load the full playbook below.
description: Plans and runs productive working meetings — deciding whether a meeting is needed at all, agendas built from outcomes with timeboxed items and pre-reads, a stated decision method and decider, facilitation techniques for divergent and convergent discussion, decision records, and follow-ups with owners and dates sent within 24 hours. Use when writing a meeting agenda, preparing to facilitate a decision or planning meeting, running a workshop or kickoff, fixing recurring meetings that waste time, or turning meeting notes into decisions and action items. Try asking
- Write an agenda for a 50-minute meeting to decide whether exports ship in Q4
- Turn these messy meeting notes into decisions and action items with owners
- Audit our recurring meetings and tell me which to cancel or make async
Use it when
- Writing an agenda for a decision or planning meeting
- Preparing to facilitate a workshop or kickoff
- Recurring meetings run long and produce nothing
- Converting meeting notes into a decision record
Skip it when
- You are running a team retrospective, use retrospectives instead
- You need to record an architecture decision for the codebase, use adr-records instead
How to check the result
The invite states an outcome, decider and timeboxed agenda, and notes sent within a day list decisions with rationale and actions with one owner and a due date each.
The common mistake it prevents
Calling a meeting to discuss a topic without naming the outcome or the decider, so it ends with an agreement to follow up and nothing decided.
A meeting is done well when it had a reason that could not be handled asynchronously, the right people came prepared, it ended with explicit decisions or outputs and owned next steps, and everyone not in the room can learn the outcome from a short written record within a day. The test of a meeting is what changed because of it.
Ground rules
- Every meeting states its outcome in the invite: “Decide X”, “Agree Y”, “Produce Z”. “Discuss X” is not an outcome.
- If the goal is to share information one way, write it instead. Meet for decisions, conflict, creative work, and relationships.
- Name the decider and the decision method before the meeting, not during it.
- Invite the smallest group that can decide: people who decide, who must commit, and who hold missing information. Others get the notes. Aim for ≤ 8 for decisions.
- Pre-reads go out ≥ 24 hours ahead; the first 5–10 minutes may be silent reading if people did not read.
- End on time. A meeting that needs more time ends, and a follow-up is scheduled with a narrower question.
Is a meeting needed?
| Purpose | Default format |
|---|---|
| Status update, FYI | Written update; no meeting |
| Simple decision with clear options | Async doc with comment deadline and named decider |
| Complex or contested decision | Meeting with pre-read |
| Brainstorm or design session | Workshop with facilitation |
| Conflict, sensitive feedback | Meeting (small, often 1:1) |
| Kickoff of multi-team work | Meeting, then async follow-ups |
Workflow
- [ ] 1. Write the outcome in one sentence; confirm a meeting is the right format
- [ ] 2. Choose participants and their roles (decider, contributors, informed)
- [ ] 3. Draft the agenda: items as questions, each with timebox, owner, desired output
- [ ] 4. Send the invite with agenda and pre-read ≥ 24 h ahead
- [ ] 5. Facilitate: open, run items, park tangents, close with decisions and actions
- [ ] 6. Send notes with decisions and actions within 24 h
- [ ] 7. Track actions to completion; review at the next touchpoint
Agenda
Write items as questions to answer, ordered by importance, with time that adds up to less than the slot (leave 5 minutes buffer per hour).
# Q4 export roadmap decision — Thu 14:00–14:50
Outcome: Decide whether background exports ship in Q4 or Q1.
Decider: Priya (Head of Product). Method: consult then decide.
Pre-read: export options doc (link) — please comment by Wed 17:00.
Attendees: Priya, Ana (eng lead), Raj (design), Mei (support), Tom (sales)
| Time | Item (question) | Owner | Output |
|---|---|---|---|
| 14:00 | Context and goal (no new info beyond the pre-read) | Ana | Shared facts |
| 14:05 | What are the costs and risks of each option? | Ana | Risk list per option |
| 14:20 | How much are customers and deals affected by Q1? | Mei, Tom | Impact estimate |
| 14:30 | Which option, and what do we give up? | Priya | Decision |
| 14:40 | Next steps and owners | Facilitator | Action list |
| 14:45 | Buffer / parking lot review | Facilitator | — |
Decision methods
| Method | When | How |
|---|---|---|
| Decider decides after consultation (default) | Most product and team decisions | Gather input, decider announces in the meeting or within 24 h |
| Consent | Team working agreements, process changes | Proposal adopted unless someone has a reasoned objection (“safe enough to try”) |
| Consensus | High-commitment, small group, time available | Everyone can support it; costly, use sparingly |
| Vote | Many reasonable options, low stakes | Majority or dot voting; tie-break by decider |
| Delegate | Experts closest to the problem | Hand the decision to them with constraints |
Use a fist-of-five check (0–5 fingers) to test support quickly: any 0–2 gets heard before the decision is final. Record dissent in the decision record; disagree-and-commit only works when the disagreement was voiced.
Facilitation techniques
| Moment | Technique |
|---|---|
| Opening | State outcome, timebox, decider, and how the parking lot works |
| Need many ideas | 1-2-4-all: alone 2 min, pairs 4 min, fours 5 min, then whole group |
| Quiet voices | Silent writing before discussion; round-robin turns |
| Dominant voice | “Let’s hear from people who haven’t spoken yet”; timebox each speaker |
| Tangent | “Good point — parking lot; we’ll assign it at the end” |
| Going in circles | Restate the question; summarize positions; ask what new information would change minds |
| Converging | Group ideas, define criteria, dot vote, then decider chooses |
| Disagreement on facts | Name the fact to verify, assign someone, move on |
| Disagreement on values | Make the trade-off explicit: “We’re choosing speed over completeness” |
| Over time | Stop; decide what can be decided; schedule a narrower follow-up |
Hybrid meetings: one person per room, or everyone joins on their own laptop; use the chat for queue order; the facilitator checks remote participants first at each item.
Decision record and notes
Send within 24 hours. Decisions and actions at the top; discussion summary only if useful.
# Notes: Q4 export roadmap decision — 2026-10-22
Attendees: Priya, Ana, Raj, Mei, Tom Facilitator: Ana
## Decisions
- D1: Background exports ship in Q4 behind a flag; scheduled exports move to Q1.
Decider: Priya. Rationale: 118 accounts blocked at month-end, $1.9M ARR exposed.
Trade-off accepted: column customization slips to Q1. Dissent: Tom (wanted scheduled
exports for two open deals); sales will offer manual workaround.
## Actions
| # | Action | Owner | Due |
|---|---|---|---|
| A1 | Update roadmap and notify sales of Q1 scheduled exports | Priya | Oct 24 |
| A2 | Split background export epic into stories | Ana | Oct 27 |
| A3 | Draft support macro for the interim workaround | Mei | Oct 25 |
## Parking lot
- Pricing of export add-on for enterprise — Priya to raise at pricing review (Nov 3)
## Open questions
- Does legal need to review export retention? Owner: Ana, due Oct 29
Action rule: verb + deliverable + one owner (a person, not a team) + date. “Look into exports” fails all four.
Recurring meetings
- Audit quarterly: for each recurring meeting, what decisions or outputs came out of the last four? If none, cancel it or make it async.
- Status meetings become written updates plus a short exceptions-only call.
- Give every recurring meeting a standing agenda and a rotating facilitator.
- Default lengths of 25 or 50 minutes leave gaps between back-to-back meetings.
Gotchas
- No stated outcome. The meeting becomes a discussion that ends with “let’s follow up”.
- Decider absent or unknown. The meeting produces recommendations nobody can act on. Reschedule if the decider cannot attend.
- Deciding in the meeting what needed data. Assign the fact-finding, decide async.
- Notes as transcripts. Nobody reads them. Lead with decisions and actions.
- Actions owned by “the team”. Nobody does them. One named person each.
- Re-litigating decided items. Reopen only with new information, through the decider.
- Too many attendees “for visibility”. Send notes instead.
Verification
- The invite has an outcome sentence, decider, method, agenda with timeboxes, and pre-read link.
- Agenda items are questions with an expected output, and total time ≤ slot minus buffer.
- Notes went out within 24 hours with decisions (including rationale, trade-offs, dissent) and actions with one owner and a date each.
- Parking lot items each have an owner or were dropped explicitly.
Output contract
# <Meeting name> — <date, time, length>
Outcome: <decide/agree/produce ...> Decider: <name> Method: <method>
Pre-read: <link, comment deadline> Attendees: <names and roles>
| Time | Item (question) | Owner | Output |
--- after the meeting ---
## Decisions (with rationale, trade-off, dissent)
## Actions | # | Action | Owner | Due |
## Parking lot (owner each)
## Open questions (owner, due) About this skill
Written and maintained by the Claude Skills editors to the same editorial standard as every skill in the directory: a selective trigger, a real workflow, known pitfalls and a verification step. First published , last updated . MIT licensed.