customer-discovery-week - SKILL.md Agent Skill

name: customer-discovery-week description: Use when the user is planning a customer discovery sprint to validate a sharp problem. Triggers on phrases like "how do I talk to customers", "plan a discovery week", "what should I ask", "interview script", "validate this problem", "discovery sprint", "customer research". Implements the 3-Step 1-Week Continuous Discovery process from Ch 1 of Building Rocketships.

Customer Discovery Week

What this does

Runs Ch 1's 3-Step 1-Week Continuous Discovery process. Returns a full discovery plan: target interview cohort, two question scripts (motivation

workflow), a survey draft, a synthesis template, and a time budget for the week. Refuses to recommend "discovery sprints" that skip interviews in favor of surveys alone — the book's anti-pattern.

When to use

Trigger when the user:

Wants to validate a sharp problem (pairs with sharp-problem-test)
Is pre-product, pre-PMF, or considering a new market
Asks for an interview script or discovery plan
Has a hypothesis and needs a structured way to test it in a week

When NOT to use

Problem is already validated and user is moving to execution — run aha-mapper or slc-or-mvp instead
User wants to build a permanent feedback system — that's listening-machine
User is asking about quantitative metrics / North Star — that's vmsoop-strategy

The 3-step process (from Ch 1)

Step 1 — Motivation interviews (5–10 calls)

Open-ended. Find out what the customer is trying to achieve, why it matters now, and what their alternatives are. Avoid leading questions. The goal is to understand the job behind the problem.

Key questions:

Walk me through the last time you tried to solve [problem area]
What did you try? Why?
What was the outcome? How did it feel?
What's the cost of NOT solving this?
If a magic solution appeared, what would it do for you?

Step 2 — Workflow interviews (5–10 calls)

Specific. Walk through the current workflow step by step. Identify the target workflow (the one your product would replace), the preceding workflow (what triggers it), and the succeeding workflow (what comes after). The book's value-chain lens.

Key questions:

Show me how you do this today, screen-share if possible
What happens right before you start this?
What do you do right after you finish?
Where do you switch tools / context / role?
What's the most painful step? Why?

Step 3 — Survey (20–80 responses)

Quantify what the interviews surfaced. The survey is for prevalence and intensity, not for discovering new things. Short (5–8 questions), distributed to a target list (LinkedIn, community, customer base).

Key questions to include:

How often does this happen? (frequency anchor)
What did you try? (alternatives confirmation)
How important is solving it on a 1–10 scale? (intensity)
Would you pay to solve it? At roughly what level? (willingness)

The synthesis template

After each step, write a one-pager:

What we heard (verbatim quotes, 5+)
What surprised us
What confirms or contradicts the hypothesis
What we need to ask in the next step

After all three steps, synthesize across into a single sharp-problem hypothesis ready to feed sharp-problem-test.

Time budget (1 week)

Days 1–2: schedule + run motivation interviews
Days 3–4: schedule + run workflow interviews
Day 5: launch survey + start synthesis
Day 6–7: collect survey results, complete synthesis, prepare verdict

Refusal mode

If the user wants to skip interviews and run a survey only, refuse and explain: surveys can quantify what interviews already surfaced, but they cannot discover what to ask. Without interviews, the survey questions encode the founder's biases, not the customer's reality.

If the user has fewer than 5 contacts available for interviews, flag — under 5 is signal you don't have access to the market, which is its own validation problem.

Inputs

The problem hypothesis to validate
(Optional) target customer description (5-dimension or thin)
(Optional) channels available for recruitment
(Optional) any prior research already done

Output structure

## Discovery plan verdict: [READY | NEEDS_MORE_INPUT]

### Target cohort
[Specific role + segment, sample size 5-10 for interviews, 20-80 for survey]

### Motivation interview script
[5-8 open questions in the order to ask them]

### Workflow interview script
[5-8 questions including the value-chain framing]

### Survey draft
[5-8 quantitative questions with answer choices]

### Synthesis template
[The one-pager structure to fill after each step]

### Time budget
[Day-by-day breakdown for 1 week]

### Recommendation
[Concrete next step — typically 'start scheduling day-1 interviews']

### Analogous case from the book
[A named example]

Worked example from the book

Calendly's pre-launch discovery (Ch 1)

Cohort: 10 sales reps + 5 recruiters who scheduled 5+ external meetings per week
Motivation: "What was the worst meeting-scheduling experience you've had recently?" — surfaced lost-pipeline stories from sales, no-show patterns from recruiting
Workflow: shared screen during a typical scheduling pass — exposed the back-and-forth email loop and the spreadsheet-of-slots workaround
Survey: 60 responses confirming daily frequency and quantified time cost (~5 hours/week per AE)
Synthesis: ready hypothesis — "scheduling external meetings is a daily, painful, role-specific problem with a real cost"

Gotchas

Stop hunting for new themes after interview 8–10. If the next three interviews all confirm the same picture, you have enough.
Survey before interviews = noise. The survey asks what interviews already told you. Reverse the order and the survey is the founder's hypothesis with extra steps.
Recruit from your real target, not your network. Friends in adjacent roles will tell you what you want to hear.
Record the interviews (with consent) and quote verbatim. The exact words matter — they're how you'll write copy that converts.

Anti-patterns to flag

If the user's plan has any of these, surface them:

"We'll send a survey and skip the interviews" — refuse; that's founder-bias confirmation
"We don't know any potential customers" — research-access gap; recruit before running the sprint
"We'll talk to 2 people" — too thin; need 5+ at minimum per interview step
"We'll talk to current customers about a new product idea" — they are not the right cohort for the new problem; recruit fresh

Source

Ch 1: Customer Discovery: Identifying Sharp Problems.

About this skill

This skill is part of the ProductMind Skills catalog — open-source product-judgment skills for AI coding agents.

Source: Building Rocketships by Oji and Ezinne Udezue — Chapter 1. This skill applies one named framework from the book. It works standalone in any agent that supports the SKILL.md standard.

Want the orchestrated version? The RocketshipsGPT agent runs all 13 book skills together, with persistent memory, framework-fidelity evals, voice-checked output, and the book's anti-pattern catches built in. → rocketshipsgpt.productmind.co

License: Skill content licensed under CC BY-SA 4.0. Attribution required; derivative works must share alike. See ADR 0005 for rationale.

Skill version: 1.0.0 · Published: 2026-06-05