Meeting Prep

When to use

Trigger this skill when:

• A first meeting is booked with a new account and the rep needs prep within 24 hours.
• A second meeting with new attendees (e.g. a technical evaluator added) needs incremental prep.
• A critical stage-gate meeting (exec alignment, procurement kickoff, QBR) is on the calendar.
• A renewal conversation requires context on the customer's last 12 months.

Do not use this for internal meetings or for first-touch cold prospecting — use lead-research for outbound prep.

Inputs

Required:

• meeting_context — date/time, meeting purpose, expected length.
• attendees — list of attendees with name, title, and company. Email addresses help for disambiguation.
• company_name — the customer or prospect company.

Optional but recommended:

• calendar_invite — if parseable, extracts attendees automatically.
• prior_touches — emails, call notes, Gong summaries from previous conversations.
• research_brief — if lead-research has already been run, pass its output.
• deal_stage — early / mid / late, affects the agenda shape.
• our_goal — what we want to leave the meeting with (e.g. "MSA redlines returned," "champion agrees to intro CISO").

Outputs

A one-page Markdown brief:

1. Meeting header — who, when, purpose, duration.
2. Our success metric — one sentence defining what a successful meeting looks like.
3. Company snapshot — 3-5 lines (pulled from or compressed from research_brief).
4. Attendees — per person: title, tenure, prior companies, recent LinkedIn activity, inferred priorities.
5. Likely priorities by role — for each attendee, 2-3 priorities based on their role and tenure.
6. Draft agenda — time-boxed segments with owners.
7. Talking points — 3-5 points tied to research, each with a supporting artifact to reference.
8. Anticipated objections — 2-4 objections likely to surface, with responses.
9. Risks to the meeting — what could derail it, with mitigation.
10. Actions / follow-up plan — what we intend to commit to.

Tool dependencies

• WebSearch / WebFetch — required. Pulls LinkedIn profiles and recent activity.
• Google Calendar MCP — optional but very useful; extracts attendees and timing from the invite.
• Salesforce / HubSpot MCP — optional; pulls prior deal history and notes.
• Gong / Chorus (if MCP available) — optional; pulls summaries of prior calls with this account.
• Notion MCP — optional; pulls case studies and internal playbooks to reference.

Procedure

1. Resolve the attendees. For each, confirm LinkedIn profile URL. Disambiguate common names using the company + title.
2. Pull per-attendee context:

• Tenure at current company.
• Prior companies and roles.
• Posts/articles/comments from the last 60 days.
• Relevant conference talks.
• Anything mutual — shared connections, shared past employer, shared conference.

1. Infer role priorities. Use the role + tenure heuristic:

• New in seat (3 years) → focused on incremental improvement, risk management.
• Role-specific defaults: CISO priorities differ from CRO priorities; list 2-3 per attendee.

1. Compress company snapshot. If research_brief exists, summarize to 5 lines. If not, run a mini-version of the research pattern.
2. Draft the agenda. Structure depends on meeting type:

• Discovery: 5 min intros / 20 min their context / 20 min our view / 10 min next steps.
• Demo: 5 min intros / 5 min frame / 30 min demo / 15 min Q&A / 5 min next steps.
• Exec alignment: 5 min intros / 10 min progress / 15 min decisions needed / 5 min next steps.

1. Write talking points. Each point names a trigger or insight from research, and an artifact to reference (a case study, a data point, a diagram).
2. Anticipate objections. Based on the attendees' roles and the stage: CFO will ask about ROI timeline, CISO will ask about data residency, VP Eng will ask about API rate limits. Write a short response per objection, tied to evidence.
3. Identify meeting risks. Common risks: the real decision maker won't be there; a surprise attendee from a competing team; timeline compression. For each, state a mitigation.
4. Define the success metric. One sentence. "Success = champion agrees to schedule security-team intro within 5 business days." Vague success metrics produce vague meetings.
5. List intended follow-ups. What will we commit to sending? Who owns each?

Examples

Example 1 — discovery call with Acme Retail

Inputs: Meeting booked for Thursday, 30 minutes, discovery. Attendees: Priya Shah (VP Data, Acme Retail, 6 weeks in role) and Luis Ortega (Director, Data Engineering, 2 years in role). Our goal: get invited to a follow-up working session with their head of engineering.

Output excerpt:

> Success metric: Priya agrees to a 60-minute architecture working session within the next 14 days with Luis and her head of engineering. > > Attendees > > - Priya Shah — VP Data, 6 weeks in role, ex-Warby Parker (Director, Data Platform). LinkedIn activity: commented on 2026-03-18 on a post about real-time CDPs saying "this is exactly the architecture we're building at Acme." Mutual connection: Jordan Lee, our SE, worked with her at Warby Parker. > - Luis Ortega — Director, Data Engineering, 2 years at Acme. Posted about dbt model performance 2026-02-22. Prior: 4 years at Shopify Plus partner agency. > > Likely priorities > > - Priya: (1) visible 100-day plan she can show the board; (2) hiring her team without bottlenecking roadmap; (3) defending the Snowflake-based architecture she's inheriting. > - Luis: (1) reducing pipeline toil; (2) managing dbt model sprawl; (3) not being the bottleneck during the CDP rollout. > > Talking points > > 1. Warby Parker's 90-day rollout — ref Jordan's direct experience. Supporting: the teardown deck. > 2. Managed pipeline vs. in-house — specifically addresses Luis's hiring gap. Supporting: the staffing model doc. > 3. Real-time latency benchmarks on Shopify + Snowflake stacks — 45s median. Supporting: the benchmark PDF. > > Anticipated objections > > - "We already have Segment." → Acknowledge; reframe around activation latency and identity resolution; redirect to a benchmark comparison. > - "We'd rather build this ourselves." → Acknowledge; reframe around time-to-first-value; redirect to the shared 90-day plan at Warby Parker. > > Risks > > - Priya may bring an unannounced attendee from procurement. Mitigation: ask at the top "anyone else planning to join?" > - 30 minutes is tight. Mitigation: pre-send the one-pager and propose extending to 45.

Example 2 — late-stage exec alignment with Nimbus Logistics

Inputs: Meeting is between their new CIO and our CRO. Deal is 6 weeks into a 12-week evaluation. Success metric: CIO commits to an expedited procurement path. Output brief emphasizes the CIO's stated "consolidate visibility spend" priority, names their stalled Project44 renewal as a forcing function, and pre-empts the likely CFO ROI objection with a specific 14-month payback model.

Constraints

• The brief fits on one page (roughly 500-700 words). Longer briefs don't get read.
• Every talking point names a specific artifact to reference. No "we should bring up value."
• Every objection response follows Acknowledge → Reframe → Redirect.
• Success metric is a single sentence with a concrete verb and a date ("agrees to schedule X by [date]").
• Attendee sections cite sources for non-LinkedIn claims (e.g. conference talk URL).
• Do not invent attendee details. If LinkedIn is private, say so.
• No emojis.

Quality checks

Before returning:

• [ ] Header includes meeting time in local timezone with date.
• [ ] Success metric is a single sentence with a specific, verifiable outcome.
• [ ] Every attendee has tenure, prior company, and at least one piece of recent activity (or explicit "no recent public activity").
• [ ] Each attendee has 2-3 inferred priorities.
• [ ] Draft agenda is time-boxed and sums to the meeting length.
• [ ] Each talking point names the artifact to bring.
• [ ] At least 2 anticipated objections, each with a reframe.
• [ ] At least one meeting risk is named with a mitigation.
• [ ] Follow-up actions have named owners.
• [ ] Brief fits on one printed page.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: tahirraufkeeyu
• Source: tahirraufkeeyu/software-development-agent-stack--sdas
• License: MIT

Install and usage instructions live in the source repository linked above.