agentleFS
ajstars1ajstars1/founder-stack/customer-discovery/SKILL.md

customer-discovery

Skill1 starsChanged 17 days agoMITView on GitHub

Customer interview prep and debriefs (Blank, Constable): question guides, recruiting, commitment scoring, ledger updates. Use before any customer conversation, after one, or for \"how do I talk to customers?\"

7 other projects use a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    The dual-role trap: if the meeting is also a sales/service delivery, do the job first, learn second — then ask permission for 5 minutes of …

    line 34

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Customer Discovery & Interviewing

Method: Steve Blank & Bob Dorf (Customer Development, The Startup Owner's Manual; free course: How to Build a Startup) and Giff Constable (Talking to Humans, 12 Tips).

Foundation: there are no facts inside the building — and discovery is done by the founders, never delegated (not to an employee, a survey form, or an AI). This skill PREPARES and DEBRIEFS interviews; only real conversations produce evidence.

The interview rulebook (Constable's 12 Tips, condensed)

  1. Work to their schedule, not yours.
  2. Get psyched to hear things you don't want to hear — the goal is learning, not a sale.
  3. Disarm politeness training — people won't call your baby ugly; tell them honesty is the only way they can help you.
  4. Start with behavior, not feedback — their current process in the problem space first; your solution last (if at all). Ask about competitors, even indirect ones, and how they found them.
  5. Open-ended questions only — never yes/no; never embed the answer you want.
  6. Listen, don't talk — ~80% silence; don't fill pauses; don't sell.
  7. Encourage but don't influence — nods and "interesting", nothing that steers.
  8. Follow your nose and drill down — always chase the "why" behind the "what".
  9. Parrot back to confirm — they'll correct or refine.
  10. Ask for introductions — leave every interview with 2–3 names.
  11. Do the paired follow-ups — one person interviews, one takes notes where possible.
  12. Write notes immediately — within 30 minutes, before details fade.

Named failure modes to avoid (Blank): Death by Demo · Death by PowerPoint · Assuming You Know · Customers Lie · The Distracted Customer · The User, the Buyer & the Saboteur (B2B roles differ — interview them separately) · overweighting one anecdote.

Interview PREP protocol (what this skill produces before a conversation)

  1. Who: target archetype + where to find them. Free interview streams founders forget: existing customers at the moment of purchase, lapsed customers (information-dense — they chose to leave), referral chains ("who else do you know…?").
  2. Hypotheses under test: pull 3–5 ASSUMPTIONS from the F/A/U list; for each, the pass/fail signal.
  3. Question guide (a guide, not a script): behavior-first, past-tense stories ("walk me through the LAST time you…"), the trigger question ("what was happening that day that made you act?"), decider/payer question, price anchored to real past spending — never "would you pay X?". Localize phrasing to the customer's language.
  4. The dual-role trap: if the meeting is also a sales/service delivery, do the job first, learn second — then ask permission for 5 minutes of questions.
  5. Cadence: consumer businesses need 50–100 conversations; ~12–15/week is a serious pace. Track weekly counts by face-to-face / phone / video.

Interview DEBRIEF protocol (when the founder reports back)

  1. Capture what was said/done (their words, quoted) strictly separate from interpretation.
  2. Score commitment signals and grade the evidence E0–E4: E0 compliments/opinions (count for nothing) · E1 told stories of past behavior · E2 observed behavior · E3 commitment — money promised, real time given, actual introductions (names/numbers) · E4 transaction. "Great idea!" with no E3+ = politeness, not validation. An assumption is working-validated at E3 from ≥3 independent people; pricing claims only at E4.
  3. Propose ledger deltas explicitly (the F/A/U Belief Ledger is the single source of truth): which assumptions changed grade (A7: E1 → E3), which were contradicted, which new items appeared. n=1 is a direction, never a verdict.
  4. Watch for: the founder pitched instead of listened; positives inflated by politeness; a single story treated as a pattern.
  5. Persist the debrief to discovery/interviews/<date>-<n>.md in the project, with gaps (⬜) to close at the next touchpoint.
Interview #N — [who], [date], [type], interviewer: [founder]
WHAT THEY SAID/DID (facts only, quoted):
WHAT I THINK IT MEANS (interpretation, separate):
COMMITMENT SIGNALS (money/time/intros):
F/A/U MOVEMENTS (gained evidence / contradicted / new unknowns):
REFERRALS COLLECTED:
DID I PITCH INSTEAD OF LISTEN? (honesty check)

Every question guide is hypotheses in disguise; every debrief is bookkeeping on your beliefs. Nothing here is customer evidence until a real human says it.

Ledger discipline (v2): reference Belief Ledger IDs ([A4], [F7]) for every claim; propose new/changed items in a closing "Ledger delta" section rather than restating claims — restated claims drift.

Fill-in templates: TEMPLATES.md in this skill's folder — copy the structures into your project's discovery/ and fill with labels.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.