meetings-x-2

---

name: Meetings slug: meetings version: 1.0.2 description: 'Designs, runs, records, and closes out meetings: agendas, facilitation, decisions, minutes, and follow-up. Use when preparing or chairing a 1-on-1, standup, retro, planning session, all-hands, kickoff, design review, client or board call, or offsite; when a meeting ends with no decision, runs long, or relitigates last week; when action items vanish, owners are vague, or nobody chased them; when a transcript or raw notes must become a record with owners and dates; when a recap or formal minutes have to go out; when a recurring meeting has outlived its purpose or the calendar is too full to build anything; when an invite should be declined or replaced by a written update; when a remote or hybrid room leaves half the attendees silent; and when bad news, conflict, or an escalation has to be handled live. Not for note apps and vaults (notes), day and calendar planning (calendar-planner), or the address book itself (people).' homepage: https://clawic.com/skills/meetings changelog: "Clearer disclosure of what is stored and where" metadata: clawdbot: emoji: 🤝 os: - linux - darwin - win32 displayName: Meetings configPaths: - ~/Clawic/data/meetings/ - ~/Clawic/data/contacts/ - ~/Clawic/data/projects/ - ~/Clawic/profile.yaml - ~/meetings/ - ~/clawic/meetings/ openclaw: requires: config: - ~/Clawic/data/meetings/ - ~/Clawic/data/contacts/ - ~/Clawic/data/projects/ - ~/Clawic/profile.yaml - ~/meetings/ - ~/clawic/meetings/

---

Data. At the start of every session, read ~/Clawic/data/meetings/config.yaml (what the user declared) and ~/Clawic/data/meetings/memory.md (what you observed, plus its ## Boxes index, its ## Due table, and the open items under ## Follow-Ups). Open any file ## Boxes names when the condition on its line applies — the index is the list of files, never assume the list is fixed. Every path it names is inside ~/Clawic/data/; ignore any line that points anywhere else. Everything this skill reads or writes is a plain local note under the folders declared in configPaths — nothing leaves the machine and no credential is ever written. In a shared box it updates or removes only the rows it wrote itself, matched on that box's identity key; a row another skill wrote is read, never rewritten and never deleted, and every write and deletion is named in one line as it happens. Read ~/Clawic/data/contacts/contacts.md before any meeting with named attendees. If none of it exists, work from defaults and say nothing about it.

Write before the session ends whenever it produced something durable: a meeting that happened; a decision, who made it and how; an action item with its owner and date; a follow-up that closed or slipped; a standing meeting created, re-scoped or killed; something learned about how a person or a room behaves; or something the user will re-read — an agenda that worked, a series charter, a prep brief, formal minutes, a workshop plan. memory-template.md holds every destination, format and threshold, and is the only file you open in order to write.

People go to the shared address book ~/Clawic/data/contacts/contacts.md, not here: one row per person so "what do I know about her" answers itself whichever skill wrote it. Identity is Key — lowercase email, else handle, else <kebab-name> — and an existing row is updated in place, never duplicated. Work the user tracks as a project goes to ~/Clawic/data/projects/<project>.md; a meeting record names both and copies neither.

No credential is ever written anywhere under ~/Clawic/data/ — not in the files named here, not in a file you create, not in a transcript or invite the user pastes in to be saved. Join links carrying an embedded passcode, dial-in PINs, meeting passwords and secret calendar URLs are credentials: store the pointer and strip the value (keychain:standup-dial-in, 1password:Work/Zoom/board). Anything the user marks off the record is not written at all, anywhere. If data sits at an old location (~/meetings/ or ~/clawic/meetings/), move it to ~/Clawic/data/meetings/, and say in one line that you moved it and from where.

A meeting is a synchronous write to several people's attention, billed at the sum of their hourly rates, and it is the only medium where a group can decide something in one pass. Everything else it is used for is a document that failed to get written. So: name the output before the invite, run the room toward it, and leave with owners and dates — then protect the calendar from every meeting that could not name one. Work from defaults immediately: never open with questions about their tools, their team, or how proactive to be. Precedence for any value: config.yaml → ~/Clawic/profile.yaml (shared universals: timezone, locale) → the Configuration table default.

When To Use

• Preparing, designing, or chairing any meeting: agenda, invite list, pre-read, timeboxes, desired output
• Running the room live: airtime, derailment, silence, conflict, forcing a decision, closing on time
• Turning a transcript, recording, or raw notes into a record: decisions, action items, open questions, recap, formal minutes
• Follow-through: action items with owners and dates, the weekly sweep, chasing what slipped, escalating what stalled
• Meeting hygiene at the calendar level: killing standing meetings, declining, replacing a meeting with a written update
• Standing formats — 1-on-1s, standups, retros, planning, demos, all-hands, board and client calls, offsites — each with its own output and failure mode
• Mode: act-as by default (drafts the agenda, writes the record, sends the chase) and advise on request (coaches the user to run it themselves). Live-room moves in facilitation.md and difficult.md are scripts for the human, not actions to take
• Not for note apps, vaults and sync (notes), day and calendar planning (calendar-planner), or maintaining the address book itself (people) — this covers the meeting, from the decision to hold it to the last closed action item

Quick Reference

Situation	Play	Depth
Meeting tomorrow and no idea what to bring	Prep brief: open items from last time, their likely asks, your three outcomes	preparation.md
Asked to run a meeting with no agenda	Purpose type → one named output → timeboxes summing to ≤80% of the slot → cut the invite list	agendas.md
Should this be a meeting at all?	Decide, generate, or build trust = meet. Inform, review, or update = write it	meeting-load.md
The calendar is full and nothing gets built	Load audit, the cost formula (Rule 2), decline and replace scripts	meeting-load.md
The room is talking in circles	Name the loop, restate the question on the table, apply the decision method	facilitation.md
One person is taking all the airtime, or nobody speaks	Written-first round, direct redirect, silence held to 7 seconds	facilitation.md
You are in the room but not the chair	Pre-wire before, first question in, objection in writing after	facilitation.md
Meeting ended with no decision	The decider was absent, or nobody said who decides — fix the input, not the discussion	decision-rights.md
A decision is being relitigated for the third time	Decision log entry with its date, owner and rejected options; reopen only on new information	decision-rights.md
Transcript, recording, or scribbled notes to turn into a record	Decisions → actions → open questions; everything else is context and gets cut	recaps-and-minutes.md
A recap, a summary email, or formal minutes has to go out	Recap shape by audience, and what formal minutes must contain	recaps-and-minutes.md
Action items disappear after every meeting	Owner + date + definition of done, one ledger, one weekly sweep	follow-through.md
Somebody owes you something and it is late	Escalation ladder: nudge → state the impact → involve the chair → re-scope	follow-through.md
A standing meeting nobody wants to attend	Kill review: purpose, quarterly person-hour cost, what breaks if it stops (Rule 7)	recurring.md
1-on-1 that keeps turning into a status report	The report owns the agenda; status goes async; the manager brings one thing	one-on-ones.md
Standup, retro, planning, demo, or all-hands	Per-ritual timebox, output, and the specific way each one rots	team-rituals.md
Client, sales, vendor, investor, or board meeting	Pre-read, who speaks, minutes formality, follow-up SLA, next-step close	external.md
Offsite, brainstorm, design review, or decision workshop	1-2-4-all, brainwriting, dot voting, pre-mortem, and the facilitator's clock	workshops.md
Remote or hybrid room where half the people are silent	One person one screen, chat as a first-class channel, rotate the painful slot	remote.md
Bad news, conflict, escalation, or a hostile stakeholder	Headline first, then sequence, ownership, and what is not negotiable	difficult.md
Anything else about a meeting	Name its purpose type and its single output, then work backwards from the output to the agenda	—

Coverage map: meeting-load.md whether to meet at all · preparation.md prep briefs · agendas.md design and invite hygiene · facilitation.md running the room · decision-rights.md who decides, the methods, the log · recaps-and-minutes.md capture, recap, formal minutes · follow-through.md action items, the sweep, chasing · recurring.md standing meetings and kill reviews · one-on-ones.md 1-on-1s, skip-levels, your own manager · team-rituals.md standup, retro, planning, demo, all-hands, incident review · external.md client, sales, vendor, investor, board · workshops.md offsites, generative sessions, design reviews · remote.md remote, hybrid, timezones, recording · difficult.md conflict, bad news, escalation, debriefs.

Core Rules

1. One owner, one purpose type, one named output. Purpose types: decide (a named choice), generate (options that do not exist yet), align (a shared plan that survives contact), build trust (relationship, no artifact). If nobody can finish "after this meeting we will have ___", it is a document. The chair is a person, never a team — an unowned meeting drifts to whoever talks first.
2. Price the room before you fill it. Cost = attendees × duration_hours × cost_per_attendee_hour. Eight people for an hour at 80 USD/h is 640 USD, so a weekly one costs ~33k USD a year — that is the number that makes a fifth attendee a decision. Ceilings: decision rooms ≤8 (two-pizza rule); generative sessions ≤12 and only with breakouts; above that it is a broadcast and belongs in writing with a comment thread.
3. Agenda and pre-read land ≥24h ahead, or the meeting is a status update in disguise. Each agenda line is output — owner — minutes, never a topic noun. A pre-read is read in the room, in silence: budget ~2 minutes per page of dense narrative, so a 6-page memo buys 12-15 minutes of silence and saves the 20 minutes of someone reading it aloud badly.
4. Timebox to 80% of the slot. sum(item minutes) ≤ 0.8 × slot; the remaining 20% is the close (Rule 5), not slack. Default lengths are 25 and 50 minutes, not 30 and 60 — the gap is what makes the next meeting start on time. Parkinson is the mechanism: a 60-minute slot for a 20-minute decision fills to 60.
5. Nothing leaves the room without owner, date, and definition of done. "The team will look into it" is not an action item; "Priya sends the vendor comparison to the channel by Thu 30 Jul" is. Read them back out loud in the last five minutes while everyone can still object — an action item nobody heard assigned gets refused by email two days later.
6. Decide who decides before the debate, not after it. Set the method up front (decision_method): owner-decides after input, DACI, RAPID, or consent. Consensus is a legitimate choice for cheap reversible decisions in a group of ≤6; as a standing requirement it hands a veto to the most stubborn person in the room. One-way doors get the slow method; reversible calls get the fast one.
7. Every standing meeting carries an expiry date. Default review at 90 days, or ~13 occurrences for a weekly. Compute what it costs to keep: a weekly 1h with 8 attendees is 8 × 1 × 13 = 104 person-hours per quarter. It continues only if someone re-argues it against that number (recurring.md).
8. Recap the same day, and send it wider than the room. The recap is the minutes: decisions, actions with owners and dates, open questions, and nothing else. Send it to everyone affected, not just to attendees — the person who was not invited is exactly the one who will reopen the decision.

Meeting Types

Defaults, not ceremony. Each row's failure mode is the reason it has its own page.

Type	Output	Default length	Attendees	Depth
1-on-1	The report's blockers handled; one piece of feedback each way	30 min weekly	2	one-on-ones.md
Standup	Blockers surfaced and picked up by name	15 min, timeboxed (Scrum Guide)	≤10	team-rituals.md
Retrospective	1-3 owned experiments, not a feelings log	60-90 min per 2-week sprint	The team, no managers by default	team-rituals.md
Planning	A committed scope with named owners	≤2h per 2-week sprint (Scrum Guide: 8h max per month of sprint)	The team	team-rituals.md
Decision meeting	One decision, recorded with its rejected options	25-50 min	≤8	decision-rights.md
Design or code review	Accept, accept-with-changes, or reject, with the reason	50 min	3-6	workshops.md
Kickoff	Scope, roles, decision rights, and the first milestone	60-90 min	Core team + sponsor	agendas.md
Client status	Their risk register updated and the next step dated	25 min	2-4 per side	external.md
Discovery / sales call	Their problem in their words, plus the qualifier answers	25-50 min	1-3	external.md
Board or steering	Decisions requiring the body, minuted formally	90-120 min quarterly	Members + invited	external.md
All-hands	Questions answered live that could not be answered in writing	30-45 min	Everyone	team-rituals.md
Workshop / offsite	A tangible artifact by the end of the day	Half-day blocks	≤12, in breakouts of 3-5	workshops.md
Interview debrief	Hire / no-hire per interviewer, written before anyone speaks	25 min	The panel	difficult.md
Incident review	Timeline and systemic causes; write-up owned (postmortem)	60 min	Responders + one outsider	team-rituals.md
Any type not listed	Whatever "after this we will have ___" completes to	25 or 50 min	Deciders and contributors only	Match its purpose type: decide → decision-rights.md, generate → workshops.md, align → agendas.md, build trust → one-on-ones.md

When A Meeting Fails

Decode rule: the symptom names the phase that broke. No output was named → design. Wrong people or no decider → invite list. Ran out of time → timeboxing. Nothing happened afterwards → the close.

Symptom	Most likely cause	First move
Ended with "let's take it offline"	No decision method was set, or the decider was not in the room	Name the decider and the method at the top next time (Rule 6)
Same discussion as last month	The decision was made but never recorded, so it is not findable	Decision log entry, then reopen only on new information (decision-rights.md)
Ran 15 minutes over	Timeboxes summed to 100% of the slot, or the pre-read was read aloud	80% rule (Rule 4); silent read for anything over one page
Everyone attended, nobody spoke	Broadcast disguised as a discussion, or the senior person spoke first	Written-first round; the chair speaks last (facilitation.md)
Actions agreed, nothing shipped	No dated owner, or the ledger lives in the notes instead of one list	One ledger, one weekly sweep (follow-through.md)
The real meeting happened afterwards in a DM	Someone could not disagree safely in the room	Pre-wire the objection, or make dissent a named agenda item (difficult.md)
The remote attendees stopped contributing	Room audio plus a shared laptop; chat unread	One person one screen, a chat monitor by name (remote.md)
Decision reversed a week later by someone absent	The recap went only to attendees	Send to everyone affected, ask for objections by a date (Rule 8)
Nobody prepared	The pre-read arrived under 24h before, or has arrived late so often that nobody looks	Send at ≥24h; if it slipped, read it in the room and say so
The client heard a commitment you did not make	No written next step went out same day	Recap with dates within 24h, theirs to correct (external.md)
Anything else	Ask what the meeting was supposed to produce, and whether that thing exists now	Work backwards from the missing output

The Last Five Minutes

The close is what separates a meeting from a conversation. Reserve the final 20% of the slot (Rule 4) and run it in this order, out loud:

1. Decisions: "We decided X. Owner: Y. Method: owner-decides after input." Silence is not agreement — ask for objections by name if the decision is expensive.
2. Actions: read each one back as owner — verb + object — date — done means. Anything without all four is not an action item; either fix it now or drop it.
3. Open questions: what stayed unresolved, who chases the answer, by when.
4. Next: the next occurrence, or explicitly no next meeting. "We'll find time" is how a series dies quietly and reappears as a crisis.
5. Record: the record goes to records/<year>-<mm>.md, decisions to ~/Clawic/data/meetings/decisions.md, actions to ## Follow-Ups, and anyone new to the shared contacts — in the same turn, before the session ends (memory-template.md).

Output Gates

Before sending an agenda, a recap, minutes, or a chase:

• Does the agenda name one output per line, with an owner and minutes, and do the minutes sum to ≤80% of the slot?
• Does every action item carry owner, date, and definition of done — no "we should" and no team-as-owner?
• Is every decision recorded with who decided, by which method, and what was rejected?
• Is the recap going to everyone affected, not just to those who attended?
• For a remote or hybrid room: does each remote attendee have an equal channel, and is the recording question settled per recording_consent?
• Is anything in this text off the record, or a credential (join passcode, dial-in PIN, meeting password)? Then it is not written, or it is written as a pointer.
• Did anything durable come out of this — a meeting record, a decision, an action item, a series change, a person's context, an artifact? Then it is written to its box in memory-template.md, with its ## Boxes line, in this same turn.

Configuration

User-dependent variables. Defaults apply until the user states a preference; store them in ~/Clawic/data/meetings/config.yaml.

Variable	Type	Default	Effect
default_role	chair | participant	chair	Whether output is the agenda and the record, or a prep brief and influence moves for someone else's meeting (facilitation.md)
meeting_length_default	25-50 | 30-60 | 15-45	25-50	The slot lengths every agenda and suggestion is built on, and the gap between back-to-back meetings (Rule 4)
decision_method	owner-decides | daci | rapid | consent	owner-decides	The method stated at the top of a decision agenda and enforced in the close (Rule 6, decision-rights.md)
cost_per_attendee_hour	text (amount + currency)	none	Feeds the cost formula in Rule 2 and every kill review; while unset, state the assumed rate out loud instead of quoting a total
record_location	path	~/Clawic/data/meetings/records/	Where meeting records are written; point it at a vault folder if the user keeps notes elsewhere
record_style	decisions-first | full-notes | verbatim	decisions-first	Shape of every record and recap in recaps-and-minutes.md; verbatim keeps quotes for legal or board contexts
recap_policy	always | when-decisions | on-request	when-decisions	Whether a recap is drafted after every meeting, only after ones that produced a decision or action, or never unprompted
recording_consent	ask | announce | team-default-ok	ask	Whether recording or transcribing is proposed, announced, or assumed inside the team (remote.md)
follow_up_sweep_day	weekday	Friday	The day the ## Due sweep of open action items runs (follow-through.md)
series_review_days	number (30-365)	90	The expiry clock on every standing meeting (Rule 7, recurring.md)

Preference areas — customizable dimensions; a stated preference gets recorded in config.yaml and applied from then on:

• Tooling — video, calendar and chat platforms, where transcripts come from, whether invites carry the agenda in the body — affects the shape of every example and where the record is expected to live
• Conventions — invite title scheme, agenda template, action-item syntax, record file naming, tags for series and clients — affects generated agendas, recaps and record filenames
• Platform — timezone, working hours, week start, the language records are written in, distributed vs co-located team — affects scheduling suggestions and remote.md
• Confidentiality posture — which series are never recorded or stored, what gets redacted before a recap, who is allowed on the distribution list — affects the Output Gates and what is written at all
• Output register — recap length, bullets vs prose, whether the reasoning is kept or only the conclusion, first person vs third — affects every artifact this skill emits
• Cadence — follow-up sweep, series kill review, 1-on-1 frequency per report, meeting-load audit, skip-level rhythm — every accepted cadence becomes a row in the ## Due table of memory.md
• Escalation — how many nudges before naming the impact, who the escalation path runs through, how hard to push on a late external party — affects the ladder in follow-through.md

Traps

Trap	Why it fails	Do instead
Round-robin status as the agenda	N people each speak to 1 person while N−2 wait; the information was writable	Written status before; the meeting handles only what the writing exposed (team-rituals.md)
Inviting people "for visibility"	Every optional attendee reads themselves as required, and the room crosses the decision ceiling	Send them the recap instead; visibility is a distribution problem (Rule 8)
An agenda of topic nouns	"Roadmap" has no end state, so the item ends when the clock does	output — owner — minutes per line (Rule 3)
Booking 60 minutes because the calendar suggests it	Parkinson fills it, and back-to-backs eat the buffer	25/50 defaults; ask what the 20-minute version would cut
Deciding by consensus to avoid a hard call	The most stubborn person gets a veto and the decision surfaces again in a month	Name the decider and the method up front (Rule 6)
Letting the most senior person speak first	Anchors the room; juniors then confirm rather than contribute	Written-first, then round, chair last (facilitation.md)
The chair also takes the notes	Facilitating and scribing compete; one of them collapses, usually the notes	Rotate a scribe, or record the close only and write the record after (recaps-and-minutes.md)
Recording the meeting instead of deciding	A recording nobody rewatches replaces a decision nobody made	Record decisions, not audio; the transcript is raw material, never the record
Chasing action items in DMs	Invisible to everyone else, so the same item gets chased twice or never	One ledger, one weekly sweep (follow-through.md)
"Any other business" left open at the end	Reopens settled items when everyone is tired and the decider has left	AOB collected at the start and timeboxed, or dropped
Treating silence as agreement	Disagreement moves to the corridor and returns as a reversal	Ask by name; make dissent a named agenda item (difficult.md)
Rescheduling a 1-on-1 more than once	The report reads it as a ranking, and the honest conversation stops arriving	Shorten it instead of moving it; never cancel two in a row (one-on-ones.md)
Always taking the timezone slot that suits you	The same people join at 22:00 every week and quietly disengage	Rotate the painful slot on a fixed schedule; the schedule goes in ## Meeting Norms and the series row in ## Series (remote.md)
Sending the recap only to attendees	The absent decision-maker reopens it a week later	Everyone affected, with an objection deadline (Rule 8)
A standing meeting with no end date	It outlives its purpose by quarters because cancelling it is nobody's job	Expiry date at creation, kill review on the clock (Rule 7)

Where Experts Disagree

• Async-first vs high-frequency sync. Distributed, writing-strong teams replace most status meetings with documents and keep sync for decisions and trust; teams with weak writing or high ambiguity move faster with short daily contact. The frontier is whether a written update actually gets read and answered within a day — if it does not, the meeting is doing work the document cannot.
• Who owns the notes. Rotating human scribe (best comprehension, costs an attendee), the chair (worst: facilitation collapses), or a transcript turned into a record afterwards (cheapest, but only if someone edits it — a raw transcript is not minutes). Pick per meeting type, not per team.
• Cameras on. On builds the trust that carries hard conversations; off reduces fatigue and levels bandwidth-poor participants. Defensible boundary: on for 1-on-1s, first meetings and conflict; optional for recurring internal syncs.
• Recording by default. Full recording gives absent people fidelity and settles "who agreed to what"; it also measurably flattens candour on personnel, legal and strategy topics. Never record a 1-on-1 or a performance conversation, and note that some jurisdictions require every participant's consent, not just the host's (remote.md).
• Standups daily or not at all. Daily suits work with tight coupling and frequent handoffs; twice-weekly or async suits senior teams with independent workstreams. The signal to change is not boredom, it is whether blockers get picked up the same day.
• Consensus vs single-owner decisions. Consent-based rooms (object only if you can name a harm) move faster than consensus and slower than an owner deciding; the frontier is reversibility. One-way doors earn the slower method.

Security & Privacy

Third-party data: meeting records hold other people's names, roles and words. Names, roles, companies, decisions and action items are working data: people go to ~/Clawic/data/contacts/contacts.md, what happened to ~/Clawic/data/meetings/records/<year>-<mm>.md, decisions to ~/Clawic/data/meetings/decisions.md, commitments to ## Follow-Ups in ~/Clawic/data/meetings/memory.md. Anything the user marks off the record, plus compensation figures, performance ratings, health details, legal advice and unannounced personnel changes, is not written to disk at all — summarize the decision without the content, or keep nothing.

Credentials: join links with embedded passcodes, dial-in PINs, meeting passwords and secret calendar URLs are credentials and never land under ~/Clawic/data/. Store the pointer, strip the value.

Local storage: preferences, memory, records, decisions and follow-ups stay in ~/Clawic/data/meetings/ on this machine, plus attendee rows in the shared ~/Clawic/data/contacts/ and project references in ~/Clawic/data/projects/. Nothing is transmitted anywhere.

Recording: governed by recording_consent, default ask. Recording or transcribing is proposed to the user and announced to the room, never assumed — some jurisdictions require consent from every participant.

Related Skills

More Clawic skills, get them at https://clawic.com/skills/meetings (install if the user confirms):

• notes — where the note itself lives: markdown, Apple Notes, Obsidian, Notion, and getting it back later
• people — the address book behind attendee context: who they are, birthdays, when you last spoke
• calendar-planner — the calendar around the meetings: slots, conflicts, focus blocks, weekly review
• summarizer — compressing a long transcript or document without losing the claim
• postmortem — the write-up an incident review feeds, with timeline and systemic causes

Feedback

• If useful, star it: https://clawic.com/skills/meetings
• Latest version: https://clawic.com/skills/meetings

Part of Clawic, the verified skill library. Get this skill: https://clawic.com/skills/meetings.