Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Run high-signal customer discovery interviews using a scripted question hierarchy, behavior-over-opinion probes, and rapport techniques drawn from Portigal, Torres, Fitzpatrick (Mom Test), and Lewis Lin.
.claude/skills/borghei-customer-interview-script/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 79% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 100% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 62% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 57% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 73% | 0% |
Running a customer interview is harder than it looks. The difference between a 60-minute conversation that produces three actionable insights and one that produces zero is almost entirely method: what you ask, when you ask it, and what you do not ask. This skill is the live-interview companion to discovery/interview-synthesis/ (post-interview analysis) and discovery/identify-assumptions/ (which produces the questions you go in to test).
The script structure draws from four canonical sources: Steve Portigal's Interviewing Users (rapport, listening, "tell me about the last time"), Teresa Torres' Continuous Discovery Habits (story-based probing, weekly cadence), Rob Fitzpatrick's The Mom Test (avoiding compliments, opinions, and futures), and Lewis Lin's behavioral interviewing patterns (concrete-recent-relevant). The goal is to leave each interview with at least one story, one contradiction, and one surprise.
Before building the interview script, confirm these inputs. If any is unknown or vague, ASK — do not assume:
Stop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.
discovery/identify-assumptions/; write 2-3 must-answer questions and pick the interview type.assets/interview_script_template.md and pull type-specific prompts from assets/question_bank.md.discovery/interview-synthesis/.See references/interview-execution-playbook.md for the full method, scripts, question banks, pacing, and troubleshooting.
references/interview-execution-playbook.md -- read this before running an interview: the four-source frameworks, the verbatim 5-phase script, question banks by type, what-not-to-ask table, pacing, pair-interviewing roles, recording/consent, silence handling, workflow, troubleshooting, and success criteria.references/interviewing-methodology-guide.md -- read this for the deep method with worked examples (Portigal, Torres, Fitzpatrick, Lin).references/red-flags.md -- read this when auditing a transcript or technique for biasing anti-patterns before trusting the findings.assets/interview_script_template.md -- ready-to-customize 5-phase interview script.assets/question_bank.md -- question bank organized by interview type (problem / solution / journey / churn).In scope: live discovery interview script structure and pacing; question banks (problem / solution / journey / churn); rapport, listening, and silence techniques; recording consent and retention policy; pair-interviewing role definitions.
Out of scope: interview synthesis and theme clustering (discovery/interview-synthesis/); assumption mapping before interviews (discovery/identify-assumptions/); survey design and quantitative research; usability testing protocols (task-based, not story-based); recruiting operations (panel sourcing, screening, incentives).
Caveats: this skill produces interview technique, not research operations. Story-based interviewing is bounded by participant memory -- events older than 90 days are reconstructions. B2B enterprise interviews often need procurement-style approval; build 2-4 weeks into recruiting. Continuous discovery assumes weekly touchpoints; one-off rounds produce shallower evidence -- weight findings accordingly.
| Integration | Direction | What Flows | |-------------|-----------|------------| | discovery/identify-assumptions/ | Receives from | Top assumptions become the must-answer questions for each interview | | discovery/interview-synthesis/ | Feeds into | Transcripts and debrief notes are the input for theme clustering | | discovery/brainstorm-ideas/ | Bidirectional | Pre-interview hypotheses; post-interview ideas seeded by themes | | discovery/brainstorm-experiments/ | Feeds into | Validated pains become experiment hypotheses | | discovery/jtbd-workshop/ | Complementary | Switch interviews use this script structure as their foundation | | discovery/value-proposition-canvas/ | Feeds into | Jobs, pains, and gains captured in stories populate the Customer Profile | | execution/create-prd/ | Feeds into | Direct quotes strengthen PRD Background and Market Segments |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 20,697 | 25,159 | +22% | 1 | 1 | 0% | 2,957 | 5,351 | +81% | 0 | 0 | — |
case-02 | fail→fail | 18,212 | 22,460 | +23% | 1 | 1 | 0% | 2,633 | 4,557 | +73% | 0 | 0 | — |
case-03 | fail→fail | 16,919 | 18,900 | +12% | 1 | 1 | 0% | 2,408 | 4,308 | +79% | 0 | 0 | — |
case-04 | fail→pass | 14,855 | 18,708 | +26% | 1 | 1 | 0% | 2,306 | 4,125 | +79% | 0 | 0 | — |
case-05 | pass→pass | 15,074 | 9,463 | -37% | 1 | 1 | 0% | 2,329 | 2,824 | +21% | 0 | 0 | — |
case-10 | fail→pass | 11,082 | 12,513 | +13% | 1 | 1 | 0% | 1,590 | 3,181 | +100% | 0 | 0 | — |
case-06 | fail→pass | 14,901 | 12,833 | -14% | 1 | 1 | 0% | 1,995 | 3,225 | +62% | 0 | 0 | — |
case-07 | pass→pass | 14,395 | 11,499 | -20% | 1 | 1 | 0% | 2,149 | 2,919 | +36% | 0 | 0 | — |
case-08 | pass→pass | 11,002 | 8,405 | -24% | 1 | 1 | 0% | 1,585 | 2,555 | +61% | 0 | 0 | — |
case-09 | pass→pass | 10,263 | 4,990 | -51% | 1 | 1 | 0% | 1,547 | 2,033 | +31% | 0 | 0 | — |
case-11 | fail→pass | 13,674 | 11,268 | -18% | 1 | 1 | 0% | 1,963 | 3,086 | +57% | 0 | 0 | — |
case-12 | pass→pass | 6,936 | 4,553 | -34% | 1 | 1 | 0% | 944 | 2,008 | +113% | 0 | 0 | — |
case-13 | fail→pass | 16,265 | 20,073 | +23% | 1 | 1 | 0% | 2,459 | 4,263 | +73% | 0 | 0 | — |
case-14 | pass→pass | 8,011 | 4,756 | -41% | 1 | 1 | 0% | 1,189 | 2,221 | +87% | 0 | 0 | — |
case-15 | pass→pass | 4,875 | 3,516 | -28% | 1 | 1 | 0% | 669 | 1,842 | +175% | 0 | 0 | — |
case-16 | fail→pass | 18,991 | 23,778 | +25% | 1 | 1 | 0% | 2,937 | 4,906 | +67% | 0 | 0 | — |
case-17 | fail→pass | 12,031 | 18,655 | +55% | 1 | 1 | 0% | 1,901 | 4,018 | +111% | 0 | 0 | — |
case-18 | pass→pass | 11,552 | 9,926 | -14% | 1 | 1 | 0% | 1,811 | 2,851 | +57% | 0 | 0 | — |
case-19 | fail→pass | 17,045 | 20,614 | +21% | 1 | 1 | 0% | 2,710 | 4,407 | +63% | 0 | 0 | — |
case-20 | fail→fail | 7,319 | 6,935 | -5% | 1 | 1 | 0% | 1,118 | 2,484 | +122% | 0 | 0 | — |
case-21 | fail→pass | 15,124 | 13,367 | -12% | 1 | 1 | 0% | 2,682 | 3,409 | +27% | 0 | 0 | — |
case-22 | fail→fail | 17,563 | 22,523 | +28% | 1 | 1 | 0% | 3,118 | 5,294 | +70% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +41 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.