octavehq/meeting-prep

/octave:meeting-prep - Strategic Meeting Prep

Build a grounded, strategic prep for a specific upcoming meeting, rendered as a self-contained HTML document. Unlike /octave:brief (a static account dossier), meeting-prep is built for the conversation in front of you: it verifies who's actually in the room, establishes why this company and why now, makes the case for the product for each persona at the table, frames the winning story around their business pain, anticipates the objections and competitors you'll face, and hands you talking points and beats to run the conversation — not a word-for-word script.

The skill reads two coaching reference files at runtime:

• references/strategic-coach.md — Enterprise strategic sales coaching (ideal-customer fit, ecosystem positioning, pain-led Socratic discovery)
• references/positioning-coach.md — Product positioning coaching based on April Dunford's methodology (the positioned narrative, competitive alternatives, feature→value→emotion)

If a user replaces these files with their own coaching frameworks, the skill adapts automatically.

Key differentiators:

• vs /octave:brief — brief is a static reference dossier; meeting-prep is a coached prep tuned to one specific conversation, with a winning story and conversation beats
• vs /octave:research — research outputs plain text; meeting-prep renders a styled HTML document with grounded GTM intelligence
• vs /octave:deck — deck is a slide presentation for the audience; meeting-prep is internal prep for the seller

Ground everything — verify before you generate

This document goes into a live meeting. A single invented name, wrong title, or hallucinated "fact" destroys trust in the entire prep — and in the seller who walks in repeating it. Every person, company fact, news item, metric, quote, and reference must trace to a real source: an Octave tool result or verified web research. Never invent, never guess, never "round up" a fact.

• People are the #1 trap. Confirm each named stakeholder actually exists before putting them in the doc — resolve them (resolve_profile_from_email, enrich_person, or find_person) and link their real LinkedIn URL. If you cannot confirm a person, do not present them as fact: either omit them, or list them under a clear "⚠ Unconfirmed — verify before the call" flag. Never fabricate a contact, title, or "the seller you'll meet."
• Separate internal from customer before you flag anyone. A named "contact" or "champion" is often actually your own team — the deal owner / AE / SE. Internal people resolve to your company's domain (e.g. @<your-domain>) or show up as the CRM deal/record owner. They are the rep this prep is for, not a customer stakeholder to verify: name them in the header ("Prepared for"), and never flag a colleague with a ⚠ as if they were an unverifiable prospect — that reads as a glaring error. The CRM's synthesized "champion/primary contact" field frequently mislabels the internal rep as a customer champion; treat that field as a hypothesis to check, not a fact.
• Company & news. Link the company's website. Every news item carries a date and a source link (deep_web_research, scrape_website). No undated "recently they…" claims.
• Proof & metrics. Every reference, customer name, logo, quote, and metric comes from Octave (proof_point / reference entities or enrichment). No invented logos, no paraphrased quotes presented as verbatim, no made-up numbers.
• Link cited entities to source (internal doc, so do it). This prep is internal, so link each cited library entity (proof point, reference, persona, competitor, objection, use case, Motion ICP cell) to https://app.octavehq.com/entity/{oId} using the oId from its tool result — one click to verify. (Never put these app.octavehq.com links in a customer-facing asset.)
• When in doubt, mark it. A clearly flagged "unverified — confirm live" beats a confident fabrication every time.

On-brand styling — internal doc, so default to YOUR brand

A meeting-prep is internal seller collateral, so it should look like the sender's own brand (your workspace's company), not the target account's — don't ask whose brand. Resolve the sender with get_workspace_company and follow the kit lookup and defaults in ../shared/brand-kit-usage.md. Respect an explicit --style or brand override.

Craft it like real collateral, don't fill a template. The output should look like the sender's brand — inline their real logo, map their tokens, vary layout to fit the occasion. The html-scaffold.md is a component-pattern reference to adapt, not a fixed stylesheet to reproduce verbatim. A reader should be able to tell at a glance whose document this is. (See ../shared/presentation-principles.md → "Look like the brand, not like a template.")

Review pass (runs by default)

Run the default review pass after generating — the always-on preflight plus the visual render-and-inspect pass, per ../shared/review-pass.md. Opt out with --skip-review; the visual pass defaults off in a --research fast run.

Usage

/octave:meeting-prep <target> [--type <meeting-type>] [--style <preset>] [--research <deep|standard|fast>] [--skip-review]

Examples

/octave:meeting-prep acme.com                                    # General meeting prep
/octave:meeting-prep [email protected] --type discovery              # Discovery call prep
/octave:meeting-prep acme.com --type demo                        # Demo prep with conversation beats
/octave:meeting-prep acme.com --type executive                   # Executive meeting, business-pain framing
/octave:meeting-prep [email protected] --type follow-up              # Follow-up with prior call context
/octave:meeting-prep acme.com --type qbr --style executive-dark  # QBR prep with specific style
/octave:meeting-prep "meeting with VP Sales at Acme"             # Context-based prep

Meeting Types

| Type | Primary Focus | |------|--------------| | discovery | Pain-led discovery questions, why-now, fit — light on pitch | | demo | Why-us per persona, conversation beats, objections/landmines | | follow-up | Updated pain from prior calls, what changed, next advance | | executive | Concise snapshot, the winning story, business-pain framing | | qbr | Value delivered, expansion angles, references | | general | Balanced across all sections (default) |

Instructions

When the user runs /octave:meeting-prep:

Step 1: Understand the Context

1.1 Identify the target:

• Email address -> Person-targeted prep (enrich person + company)
• Domain -> Company-targeted prep (enrich company + find key contacts)
• LinkedIn URL -> Person-targeted prep
• Meeting description -> Extract company/people from context

1.2 Detect or ask meeting type:

If --type is given, use it. Otherwise infer first, then confirm — don't ask cold. Do a quick deal/CRM peek (get_deal_deep_dive / list_events / list_findings) and let it pre-select the likely type: no prior meetings / early stage → Discovery; a demo already happened → Follow-up; an exec or economic buyer is involved / late stage → Executive; existing customer → QBR. Present your inference as the default ("Looks like a follow-up — 3 prior calls, last was a demo. Go with that, or pick another?"). The same peek can pre-fill attendees (from CRM contacts) and a sensible duration, so Step 1 is mostly confirmation, not interrogation. If there's no deal/CRM signal, fall back to asking:

What type of meeting are you prepping for?

1. Discovery — First conversation, understanding their world
2. Demo — Showing the product, proving value
3. Follow-up — Continuing a conversation, advancing the deal
4. Executive — High-level strategic conversation
5. QBR — Quarterly business review with existing customer
6. General — Balanced prep (default)

Your choice:

1.3 Ask meeting duration:

Duration drives how much to cover and how many conversation beats to plan — not a minute-by-minute timeline.

How long is this meeting?

1. 30 minutes
2. 45 minutes
3. 60 minutes
4. 90 minutes

Your choice:

1.4 Collect user context:

Ask if the user has any prior context to incorporate:

Do you have any prior context to fold in?

1. Call transcript or recording notes
2. Email thread or meeting notes
3. My own notes / talking points
4. No prior context — use Octave intel + coaching frameworks

Paste or describe (or press Enter to skip):

If the user provides a transcript, notes, or email thread, synthesize that context alongside Octave data — and mine it for the customer's exact language about their pain and goals. If they skip, proceed with Octave intel and coaching frameworks only.

1.5 Identify attendees:

Who's attending? (names, titles, emails — or "I don't know yet")

Whatever they give you, verify it in Step 2 before it lands in the doc. If attendees are unknown, build a likely buying committee from Octave contacts — clearly labeled as inferred, not confirmed.

1.6 Read coaching reference files:

Read the reference files:

• references/strategic-coach.md — Extract: ideal-customer fit, ecosystem/enhancement positioning, pain-led Socratic discovery
• references/positioning-coach.md — Extract: the positioned narrative (status quo → problem → category → why-us → proof), feature→value→emotion, competitive alternatives, category framing, language mining
• ../shared/presentation-principles.md — the shared output rules (label every value, no tool jargon, confirmed vs hypothesized, links open in a new tab, themed scrollbars, look like the brand not a template). Mandatory for the generation step.

If the coaching files are not found, fall back to general sales coaching best practices.

Step 2: Octave Context Gathering

The research is a layered blend: live external web (deep_web_research, scrape_website) + Octave enrichment (enrich_company/qualify_company/find_person) + the Octave library (Motions / Motion ICP cells, proof points, references, competitors, objections, use cases) + CRM & conversation intel (get_deal_deep_dive, list_events, list_findings). The mix shifts by situation — a net-new prospect leans on web + enrichment + library; an existing customer adds rich CRM/findings.

Research depth — default: deep

Run a deep research pass by default. As you start, tell the user in one line what you're doing and how to dial it down, then proceed (don't wait for a reply unless they ask to change it):

"Running a deep research pass for [Company] — live web + market/segment intel, the full Octave library, and CRM/deal history. It's the most thorough mode and takes a few minutes. Want it faster? Say standard (skip the broad market scan) or fast (Octave + CRM only, no live web)."

| Mode | What it pulls | Live web | |------|---------------|----------| | deep (default) | All layers + broad deep_web_research — company news plus segment/market trends and competitor moves | broad (3-5 targeted queries) | | standard | All layers + deep_web_research for recent company news + 1-2 macro themes | focused (1-2 queries) | | fast | Octave library + enrichment + CRM/findings only | none |

Honor an explicit --research <deep|standard|fast> flag or any in-line request to switch. A clearly quick internal sync can default to standard; otherwise default deep.

Bound it — deep is thorough, not unbounded. Gather what the relevant layers offer, cap live web to the query budget above, then start writing. Once you can ground each section, generate — don't keep researching (a thorough prep that ships beats an exhaustive one that stalls). Silently skip layers with no data (a net-new prospect has no CRM/findings).

Based on the target and meeting type, use Octave MCP tools to build a complete intelligence picture. Tell the user what you're researching and why.

Call as many tools as needed to build a thorough, grounded prep. The best preps layer multiple sources — company enrichment + person verification + recent news + segment research + Motion ICP persona narratives + value props + proof points + competitive intel + conversation findings all combine into a document grounded in real data. Not every tool applies to every meeting; pick the combination that gives you the richest, most verifiable context:

• List-vs-search guidance and the common tool tables: ../shared/octave-research-toolkit.md.
• The prep's ordered tool tables — 2a verify people/company first, 2b why-this-company/why-now, 2c why-us per persona, 2d proof/objections/competitors, 2e deal state & conversation history (always attempt with a 90-day window; silently omit if empty) — plus the internal-vs-customer check: references/tool-reference.md.

Output of this step: Present a content outline to the user for approval before generating:

MEETING PREP OUTLINE: [Company/Person] — [Meeting Type]
========================================================

Target: [Company name / Person name at Company]
Meeting Type: [Discovery / Demo / Follow-up / Executive / QBR / General]
Duration: [30 / 45 / 60 / 90] minutes
Attendees: [Verified names + roles, ⚠ flag any unconfirmed]
Style: [Will be selected in Step 3]

---

SECTIONS TO INCLUDE
-------------------

1. Header — Meeting details, date, duration, verified + linked attendees
2. Snapshot — Situation + deal state + the one outcome we want (merged; no separate "deal intel")
3. Why This Company, Why Now — fit reasons, recent news (dated/linked), segment intel, similar customers
4. Stakeholders — verified people only, linked LinkedIn, deal role, what each cares about
5. Why [Product] for Each Persona — why they care + why us + top use cases, per persona at the table
6. The Winning Story — the narrative arc grounded in their business pain
7. How to Run the Conversation — talking points & beats, what to listen for, what to ask for
8. Discovery Questions — pain- and situation-led (NOT sales-process)
9. Objections & Competitors — likely objections + responses; likely competitors/alternatives + watch-outs
10. The Line — one memorable sentence

Octave Sources Used:
- Company enrichment + fit: [Company] — [segment, fit score, key reasons]
- People verified: [N confirmed / N unconfirmed]
- Recent news / segment research: [N items, dated + sourced]
- Motion ICP cells (per persona): [Persona × Segment cells pulled]
- Value props / use cases: [N]
- Proof points / references / similar customers: [N]
- Objections / competitors: [N / which]
- Deal state: [stage, compelling event] (or "new prospect — nothing on file")
- Findings: [N recent signals] (or "none found — skipped")
- User context: [Transcript / notes / none]

---

Does this look good? I can:
1. Proceed to style selection and generation
2. Add or remove sections
3. Go deeper on any area
4. Change the meeting type or emphasis

Wait for user approval before proceeding.

Step 3: Style Selection

The prep uses the same CSS variable / style preset system as the other document skills. Full preset definitions are in ../shared/style-presets.md.

Preps default to readability-optimized presets. If --style was not provided, ask:

Pick a style for your meeting prep:

1. midnight-pro     — Dark navy, white text, blue accents (default)
2. paper-minimal    — Off-white, black type, editorial simplicity
3. executive-dark   — Charcoal + gold, premium boardroom aesthetic
4. soft-light       — Warm white + sage green, calm and approachable
5. swiss-modern     — White + red accent, Bauhaus minimal
6. Use my brand     — Extract from website or provide colors
7. Match my deck    — Use the same style as an existing /octave:deck

Your choice (or press Enter for default):

| Meeting Type | Recommended Default | |--------------|-------------------| | Discovery | midnight-pro | | Demo | midnight-pro | | Follow-up | midnight-pro | | Executive | executive-dark | | QBR | executive-dark | | General | midnight-pro |

If the user selects "Use my brand," follow the brand extraction tiers in ../shared/brand-kit-usage.md. If they select "Match my deck," ask for the deck file path and extract its CSS variables.

Step 4: Generate HTML

Build a single self-contained HTML file. The prep is a scrollable reference document — not a slide deck. Natural page scroll, sticky sidebar navigation, collapsible sections, and a print-friendly layout.

Output Directory

.octave-meeting-prep/
└── <kebab-case-name>-<YYYY-MM-DD>/
    └── <name>.html

Example: /octave:meeting-prep acme.com --type discovery -> .octave-meeting-prep/acme-discovery-2026-02-27/acme-discovery.html

Make sure .octave-meeting-prep/ is ignored by your project's .gitignore (an .octave-*/ pattern covers all Octave output dirs) so generated preps don't get committed.

Meeting Type → Section Emphasis

Not all sections are equally weighted in every meeting type. The type determines emphasis:

| Meeting Type | Emphasized Sections | De-emphasized / Condensed | |--------------|-------------------|---------------------------| | Discovery | Why This Company, Discovery Questions, Why-Us per Persona (light) | The Winning Story (lighter), Objections (lighter) | | Demo | Why-Us per Persona, How to Run the Conversation, Objections & Competitors | Discovery Questions (lighter) | | Follow-up | Snapshot (updated), The Winning Story, How to Run the Conversation | Why This Company (condensed) | | Executive | Snapshot, The Winning Story, Why This Company, The Line | Discovery Questions (fewer, strategic) | | QBR | Snapshot, Why-Us per Persona (expansion), Objections (renewal risks) | Discovery Questions (expansion-focused) | | General | All sections at equal weight | None |

Document Sections (10 total)

The full specification for each section — Header, Snapshot, Why This Company/Why Now, Stakeholders, Why-[Product]-per-Persona, The Winning Story, How to Run the Conversation, Discovery Questions, Objections & Competitors, and The Line — is in references/prep-sections.md. Read it before generating; it carries the grounding rules per section (verified people only, dated/linked news, real references, confirmed-vs-speculative competition).

HTML Architecture

Build on the shared scaffold in ../shared/doc-scaffold.md; prep-specific components (snapshot strip, persona blocks, story beats, the run-the-conversation beats, objection/competitor rows, status tags) are in html-scaffold.md. Treat them as patterns to adapt, not a fixed stylesheet: drive the palette, type, and logo from the brand kit so the output looks like the sender's real collateral. Required chrome regardless of brand: every link target="_blank" rel="noopener noreferrer", and themed scrollbars (never the bare default OS scrollbar on a dark surface).

Content Density Guidelines

Preps are reference documents — thorough but scannable:

| Section | Content Limit | |---------|--------------| | Snapshot | 2-3 sentence situation + a deal strip + 1 outcome line | | Why This Company | 3-5 fit reasons, 2-4 dated news items, 2-3 similar customers | | Stakeholders | 4-6 stakeholder cards max | | Why-Us per Persona | up to 3 personas, each: why-they-care + why-us + 2-3 use cases | | The Winning Story | 5 beats / one short paragraph | | How to Run the Conversation | 4-6 beats, each with listen-for + ask-for | | Discovery Questions | 8-12 questions max | | Objections & Competitors | 3-5 objections, 2-4 competitors/alternatives, 1-2 watch-outs | | The Line | 1 sentence |

If a section would exceed its limit, prioritize by relevance to the meeting type and trim the rest.

Step 5: Delivery

After generating the HTML file:

1. Open the prep in the default browser
2. Present a summary:

MEETING PREP READY
==================

Folder: .octave-meeting-prep/<name>-<date>/
File:   .octave-meeting-prep/<name>-<date>/<name>.html
Style:  [Preset name or "Custom Brand"]
Grounding: [N people verified, N news items sourced, N unconfirmed flagged]
Sections: [List of included sections]

Navigation:
- Scroll naturally to read through sections
- Click nav dots on the right edge to jump to sections
- Click section headers to collapse/expand
- PDF (recommended): bash "${CLAUDE_PLUGIN_ROOT:-.}"/scripts/export-pdf.sh .octave-meeting-prep/<name>-<date>/<name>.html  — or Cmd+P / Ctrl+P -> Save as PDF

---

Want me to:
1. Adjust or expand a section
2. Verify or add stakeholders
3. Go deeper on any topic (why-us per persona, the winning story, objections)
4. Change the style
5. Export as PDF (print dialog)
6. Generate a brief for this account (/octave:brief)
7. Build a competitive battlecard (/octave:battlecard)
8. Build a presentation from this (/octave:deck)
9. Done

MCP Tools Used

Common research, library, signals, and generation tools: see ../shared/octave-research-toolkit.md. Meeting-prep-specific additions:

Verification & Live Research

• resolve_profile_from_email / resolve_email_from_profile — Confirm a person exists and capture their LinkedIn
• deep_web_research — Recent news (company-level) and segment/market research, dated and sourced
• scrape_website — Verified, linkable facts from the company's own site
• get_external_brand_logo — Company logo for the header / domain check
• get_workspace_company — Resolve the sender's own company for brand styling

Competitive & Deal Intelligence

• get_competitive_insights — Where competitors win/lose, differentiation
• get_deal_deep_dive / list_deal_health — Deal state for the Snapshot

Error Handling

Standard responses (company not found, no matching Motion ICP cell): see ../shared/octave-research-toolkit.md → Standard error handling. Meeting-prep-specific:

No user context provided:

No prior context provided. I'll build the prep from Octave intelligence, live research, and coaching frameworks.

The prep will be strong on grounded intel and positioning. After the meeting, run this again with your notes for a sharper follow-up prep.

Person or attendee can't be verified:

I couldn't confirm [name] as a real contact at [company]. Rather than guess, I've flagged them as "⚠ Unconfirmed — verify before the call" (or left them out).

If you have their email or LinkedIn, share it and I'll verify and enrich them properly.

Coaching reference files not found:

Coaching reference files not found in references/. Using general sales coaching best practices.

To customize coaching frameworks, add strategic-coach.md and positioning-coach.md to the skills/meeting-prep/references/ directory.

Octave Connection Failed:

Could not connect to your Octave workspace.

I'll build the prep from your provided context and coaching frameworks. The result will focus on the winning story, conversation beats, and discovery questions without enrichment data — and I'll flag anything I can't verify.

To reconnect: check your MCP configuration or run /octave:workspace status

No Findings / Deal Data:

No conversation signals or deal record found for [company/person] in the last 90 days.

The Snapshot will present this as a new prospect and flag what to uncover. The prep leans on enrichment, live research, positioning, and your provided context.

Related Skills

• /octave:brief — Static internal account dossier (reference doc, no coaching frameworks)
• /octave:research — Deep-dive research on a company or person
• /octave:battlecard — Competitive intelligence and displacement strategy
• /octave:deck — Full slide presentation for the audience
• /octave:one-pager — Customer-facing leave-behind document
• /octave:pipeline — Deal-level coaching and pipeline strategy
• /octave:abm — Account-based planning with stakeholder mapping