validate-interview
Skill7 starsChanged 17 days agoMITView on GitHub
The Interview Kit. Writes a Mom Test style interview script about the past, not opinions about the idea, plus an outreach draft for each prospect built from what that person actually said. The user sends them. Use when the user asks for "interview questions", "customer interview script", "how do I reach out", "outreach message", or "customer discovery".
In this repository10 other files it gives its agents.
- validate-clusterSkill
- validate-competitorsSkill
- validate-ideaSkill
- validate-mineSkill
- validate-painSkill
- validate-prospectsSkill
- validate-segmentSkill
- validate-smokeSkill
- validate-verdictSkill
- validate-wtpSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 7 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 26
Before anything else, rerun /validate-prospects so the list is fresh and honours validator/do_not_contact.txt.
- Very specific: line 30
If segment.md is missing, use the target in idea.json and say so.
- Distinctive: line 36
Follow the Mom Test rules (from Rob Fitzpatrick's book of that name): ask about their life and their past, never about your idea or their …
- Distinctive: line 40
Write 8 to 12 questions in this order, worded for this problem and this group:
- Very specific: line 65
Add the privacy rules to the top of the template: ask before recording a call, keep notes in validator/interviews/, keep validator/ out of …
- Very specific: line 75
Use their quote only as it appears in prospects.md.
- Distinctive: line 80
Ask for one thing: 15 minutes to hear how they handle it today.
- Distinctive: line 84
Keep it under 80 words. Write each one for that person, about what they actually said.
- Distinctive: line 87
Add a line per draft on how to reach the person on their platform:
- Very specific: line 103
Stop the moment anyone asks not to be contacted, add them to validator/do_not_contact.txt, and delete their draft.
- Very specific: line 108
When the user brings back notes (pasted, or files in validator/interviews/):
- Distinctive: line 120
Write a message that pretends to be from someone else, or hides that the user is researching a product.
- Distinctive: line 122
Use a prospect's words as a testimonial.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
validate-interview
Public complaints tell you what hurts. Interviews tell you what people actually do about it, what it costs them, and whether they will commit. This skill writes the script and the first message to each prospect. It never sends anything. The user does that, by hand, one at a time.
What the user typed after /validate-interview: $ARGUMENTS
What it reads
validator/prospects.md (who), validator/pain.md (the pains to ask about),
validator/segment.md (the beachhead), validator/wtp.md (what is known about
price) and validator/idea.json (the hypothesis and the real-world bar).
Before anything else, rerun /validate-prospects so the list is fresh and
honours validator/do_not_contact.txt. Never draft for anyone from an old copy
of the list.
If segment.md is missing, use the target in idea.json and say so. If
pain.md shows the evidence bar failed, say that plainly: the interviews are
now the main test of the idea, so treat them that way.
Step 1: the script, written to validator/interview.md
Follow the Mom Test rules (from Rob Fitzpatrick's book of that name): ask about their life and their past, never about your idea or their future. People are polite about ideas and honest about last Tuesday.
Write 8 to 12 questions in this order, worded for this problem and this group:
- Context. "Walk me through how you handle
<the area>today." - The last time. "Tell me about the last time
<the top pain>happened." Then "what happened next?" and "what did it cost you?" in time, money or a lost deal. - What they tried. "What have you tried to fix it?" "What did you pay for
that?" "Why did you stop?" Use the tools from
triedinthemes.json. - How much it matters. "Where does this rank among the problems in your week?" "Who else is affected?"
- The money, last. "What do you spend on
<the area>today, in tools and in time?" Never "would you pay for X". - The commitment ask, at the very end. Something that costs them a little:
joining a waitlist, a second call to try an early version, an introduction
to a colleague, or a refundable preorder if
/validate-smokeset one up.
Then list the questions to avoid, so the user catches themselves: "would you use...", "do you like the idea...", "how much would you pay for...", and any question that describes the product.
End interview.md with a notes template for each call: date, person (their
handle, not a real name, unless they offer it), their exact words on the pain,
how they handle it today, what it costs, what they already pay, and the
commitment they gave (none, time, an intro, money).
Add the privacy rules to the top of the template: ask before recording a call,
keep notes in validator/interviews/, keep validator/ out of any public git
repository, and delete the notes when the validation is over or within 90 days.
Step 2: one draft per prospect, written to validator/outreach.md
For each person the user picks from prospects.md (suggest 5 to 10 who fit
the target, and say so if none do):
- Open with their own post: link it and name the specific thing they said.
Use their quote only as it appears in
prospects.md. - Say who the user is in one honest line: their real name, that they are researching this problem, and that they may build something for it ("I'm exploring whether to build something for this"). People deserve to know why they are being asked.
- Ask for one thing: 15 minutes to hear how they handle it today.
- No pitch, no product name, no link to a product, and no claim that anything exists yet.
- Give an easy way out: "no worries at all if not".
- Keep it under 80 words. Write each one for that person, about what they actually said. It is a message to a human, not a campaign.
Add a line per draft on how to reach the person on their platform:
- Reddit: a chat or private message to the handle, or a public reply in the thread if the post is under 30 days old and the subreddit allows it. Read the subreddit's rules first (with web tools if you have them, otherwise list it as a step for the user). Many ban research requests in comments.
- Hacker News: there are no private messages. Use contact details only if the person put them in their own HN profile. Never go looking elsewhere for an email address.
- X and other imports: a reply, or a DM if their DMs are open.
Tell the user the sending rules:
- By hand, at most 5 a day.
- One message per person. No follow-up if they do not answer: silence is an answer.
- Stop the moment anyone asks not to be contacted, add them to
validator/do_not_contact.txt, and delete their draft.
Step 3: after the interviews
When the user brings back notes (pasted, or files in validator/interviews/):
- Count the commitments by type, and compare them with the interview part of
bar.real_worldinidea.json(for example "30% say they would pay"). - Pull out what surprised: pains the mined data missed, prices people named, tools they use.
- Quote people only with their permission, and only as the user's notes have it, marked as interview notes, not as verified posts.
Never
- Send, post, like, follow or DM anything yourself.
- Write a message that pretends to be from someone else, or hides that the user is researching a product.
- Use a prospect's words as a testimonial.
Output
validator/interview.md, validator/outreach.md.
Next: /validate-smoke puts a page and a price in front of real people.
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
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.

