Skip to content
pm-skills

Using the Meeting Skills Family

The Meeting Skills Family is a coordinated set of five foundation-meeting-* skills for the meeting lifecycle: pre-meeting prep, post-meeting summarization, cross-meeting analysis, and outward stakeholder communication. This guide walks you through the family as a PM would encounter it in practice.

Read the family contract if you need the formal specification. This guide is for getting started.

The five skills at a glance

graph LR
    Brief[/meeting-brief<br/>private strategic prep/]
    Agenda[/meeting-agenda<br/>attendee-facing structure/]
    Meeting[The meeting happens]
    Recap[/meeting-recap<br/>post-meeting summary/]
    Update[/stakeholder-update<br/>async comms/]
    Synthesize[/meeting-synthesize<br/>cross-meeting analysis/]

    Brief --> Meeting
    Agenda --> Meeting
    Meeting --> Recap
    Recap --> Update
    Recap --> Synthesize

    style Brief fill:#e1f5ff
    style Agenda fill:#e1f5ff
    style Meeting fill:#ffe1e1
    style Recap fill:#e1ffe1
    style Update fill:#fff4e1
    style Synthesize fill:#f0e1ff
SkillPhaseAudienceKey output
/meeting-agendaPre-meetingAttendeesTime-boxed topics, owners, prep
/meeting-briefPre-meetingYou (private)Stakeholder reads, messages, asks
/meeting-recapPost-meetingAttendeesTopic summary, decisions, actions
/meeting-synthesizeCross-meetingYou / leadershipThemes, contradictions, follow-ups
/stakeholder-updatePost-meetingNon-attendeesChannel-tailored outward comms

Your first meeting with pm-skills

Imagine you have a cross-functional decision meeting Thursday afternoon. Four attendees. You’re chairing. You have 30 minutes.

Step 1. Write the agenda (2 minutes of your time)

Terminal window
/meeting-agenda "Thursday 2pm EST cross-functional sync on Q2 scope; 4 attendees; 45 min"

The skill will:

  1. Run the anti-meeting check. “does this actually need to be a meeting?” If your objective can be handled async, it’ll recommend that instead. Override if needed.
  2. Present an inference summary. “Inferred: meeting type = decision-making, duration = 45 min (as specified), meeting objective inferred from topic.” Say go to accept, or correct any value.
  3. Produce an agenda with type-tagged topics, time allocations, owners, and attendee prep.

Copy the shareable summary at the top of the output into your meeting invite. Done.

Step 2. (Optional) Private brief (3 minutes)

For the CFO-is-present decision meeting, you want private tactical prep. Run:

Terminal window
/meeting-brief @notes.md @prior-recap.md "Q2 scope meeting; CFO will push on capacity"

The skill reads your context, pulls in related prior recaps (auto-discovered via filename-chaining when available), and produces a brief with:

  • Stakeholder reads (position, stakes, concerns, tactical notes)
  • Ranked desired outcomes (must / should / nice)
  • Key messages phrased for delivery
  • Anticipated Q&A
  • Specific asks by name

The brief is private by default (visibility: private in frontmatter). It’s your prep, not a shared artifact.

Step 3. The meeting happens

Run the meeting. Take live notes or let a transcription tool (Krisp, Otter, Fireflies, Zoom) capture it.

Step 4. Recap the meeting (5 minutes of your time)

Terminal window
/meeting-recap @transcript.txt

The skill:

  1. Auto-discovers the agenda from Step 1 (filename prefix match)
  2. Uses the agenda topic list as scaffold for the recap
  3. Extracts decisions (bold-flagged), actions (with owner + due date), open questions
  4. Reconciles what was planned vs. what actually happened
  5. Produces a consolidated action-by-owner view at the end

Fabrication prohibition: if an action lacks an explicit owner in the transcript, the recap tags it [owner: unassigned] rather than guessing. If more than 30% of actions lack owners, a ⚠ Ownership reconciliation required section surfaces near the top.

Copy the shareable summary into Slack for the attendees. Post the full recap to the meeting’s doc.

Step 5. Stakeholder update (2 minutes)

For non-attendees (engineering leads, leadership) who need to know outcomes:

Terminal window
/stakeholder-update @recap.md --channel=slack --audience=engineering

Or:

Terminal window
/stakeholder-update @recap.md --channel=email --audience=leadership --cta="confirm capacity impact by Friday"

The skill:

  1. Reads the recap, extracts outcomes relevant to the target audience
  2. Detects thread continuation (prior updates on the same project auto-linked)
  3. Translates technical-to-business language as needed for the audience variant
  4. Produces a channel-tailored message inside a ## Shareable update section. copy only that section, everything below is internal audit notes

Step 6. Later: cross-meeting synthesis

After your Q2 initiative has spanned 5-8 meetings over a few months, run:

Terminal window
/meeting-synthesize @recap1.md @recap2.md @recap3.md @recap4.md @recap5.md

Or narrow by filter:

Terminal window
/meeting-synthesize --quarter=2026-Q2 --format=board-prep

Produces a cross-meeting artifact with:

  • Plain-text timeline (chronological key events)
  • Themes with confidence markers
  • Stakeholder position tracking (how views evolved)
  • Consolidated decision list
  • Decision evolution (resolved) vs. unresolved contradictions (flagged with ⚠)
  • Prioritized follow-up suggestions (High / Medium / Low)

Use for board prep, onboarding new team members, or project retrospective input.

The go-mode pattern

Every meeting skill follows the same behavioral contract: read → infer → present summary → accept go or corrections.

graph TD
    A[Invoke skill with inputs] --> Z{--go flag set?}
    Z -->|yes, skip checkpoint| B
    Z -->|no, default flow| B[Read all inputs]
    B --> C[Run inference on missing fields]
    C --> D{Invoked with --go?}
    D -->|yes| H[Produce output directly]
    D -->|no| E{Confidence on<br/>load-bearing fields?}
    E -->|all high| F[Present summary,<br/>no warnings]
    E -->|any below-high<br/>on load-bearing| G[Present summary<br/>with ⚠ flags]
    F --> I{User response}
    G --> I
    I -->|go| H
    I -->|correction| C

    style H fill:#c8e6c9
    style G fill:#fff9c4
    style Z fill:#e1f5ff

Load-bearing inference gates: stakeholder positions, primary ask, and decision-maker attribution trigger a ⚠ flag when inferred below-high confidence. You can still say go, but the flag ensures you see the risk before producing tactical guidance on weak inferences.

Explicit go-mode: pass --go to skip the checkpoint entirely. Useful when you trust the defaults or want a one-shot artifact.

When the skills chain

The skills are designed to discover each other via filenames. Here’s the chain in practice:

sequenceDiagram
    participant You
    participant Brief as meeting-brief
    participant Agenda as meeting-agenda
    participant Recap as meeting-recap
    participant Update as stakeholder-update
    participant Sync as meeting-synthesize

    You->>Agenda: /meeting-agenda "Q2 scope"
    Agenda-->>You: 2026-04-17_14-00EST_q2-scope_agenda.md

    You->>Brief: /meeting-brief @prior-recap.md
    Brief-->>You: 2026-04-17_14-00EST_q2-scope_brief.md

    Note over You: meeting happens

    You->>Recap: /meeting-recap @transcript.txt
    Recap->>Agenda: auto-discover agenda
    Agenda-->>Recap: topic skeleton + desired outcomes
    Recap-->>You: 2026-04-17_14-00EST_q2-scope_recap.md

    You->>Update: /stakeholder-update @recap.md --channel=slack
    Update->>Recap: read outcomes
    Recap-->>Update: decisions, actions
    Update-->>You: 2026-04-17_14-00EST_q2-scope_stakeholder-update-slack-engineering.md

    Note over You: months later, after 5+ meetings

    You->>Sync: /meeting-synthesize @recap*.md
    Sync-->>You: 2026-Q2_q2-scope_synthesis.md

Chaining is filename-based. no separate IDs, no lookup tables. The filename itself is the identifier, and skills cross-reference each other via the related_* frontmatter fields and auto-discovery scans.

File naming: quick reference

# Single-meeting artifacts
YYYY-MM-DD_HH-MMtimezone_title_{agenda|brief|recap}[_prepared-YYYY-MM-DD].md


# Stakeholder-update (variant suffix)
YYYY-MM-DD_HH-MMtimezone_title_stakeholder-update-{channel}-{audience}[_prepared-YYYY-MM-DD].md


# Synthesis
YYYY-Qn_topic_synthesis.md
YYYY-MM-DD-to-YYYY-MM-DD_topic_synthesis.md

Examples:

  • 2026-04-17_14-00EST_search-feature-kickoff_agenda.md
  • 2026-04-17_14-00EST_search-feature-kickoff_stakeholder-update-email-leadership.md
  • 2026-Q1_search-feature_synthesis.md

Common pitfalls

“Why does the anti-meeting check keep triggering?” (v1.1.0)

The check now requires a positive synchronous-value statement. Restate your meeting’s purpose in terms of: tradeoff to discuss, conflict to resolve, co-creation, relationship-building, or blocker escalation. If none apply, the async alternative is genuinely the better path.

“The inferred meeting type is wrong.”

The inference summary is a checkpoint for exactly this. Say meeting type is actually {correct-type} and the skill will re-infer dependent values (duration, variant behavior) from the correction.

“I have 60% of actions without owners. the recap is unusable.”

The v1.1.0 recap surfaces a ⚠ Ownership reconciliation required section when unassigned ratio exceeds 30%. Use it as your action plan: the section lists each unassigned action and suggests a follow-up mechanism (15-min sync, Slack thread, async survey). Don’t ship the recap until you’ve closed the reconciliation.

“I accidentally pasted the whole stakeholder-update including the audit notes.”

The ## Shareable update section (v1.1.0) marks the copy-safe boundary explicitly. Everything below the > [End of Shareable update section...] marker is internal. Copy only above the marker.