# Validation Test

> Test the assumption that would end the venture, with the pass bar fixed beforehand.

- Skill: `tinyhumansai/validation-test` (Agent Skill)
- Install (CLI): `npx skillmds@latest add tinyhumansai/validation-test`
- Raw SKILL.md: https://api.skillmd.com/api/skills/tinyhumansai/validation-test/raw
- Safety review: pending (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: tinyhumansai (https://skillmd.com/u/tinyhumansai)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/tinyhumansai/validation-test

---


# Validation Test

Test the thing that would kill the venture, not the thing that is easiest to
test. Those are almost never the same, and picking the second is how a studio
spends a quarter learning nothing.

## When to use

- A venture is at thesis or validating stage.
- Somebody wants to start building.

## Steps

1. **Name the riskiest assumption.** Usually demand, occasionally a supply
   constraint, rarely the technology — though the technology is what everybody
   wants to test because it is the enjoyable part.
2. **Fix the pass bar before running,** on the `validations` row. Afterwards,
   any result reads as encouraging: this is the single most important step here.
3. **Pick the cheapest method that could actually fail.** A landing page, a
   concierge build, a pre-sale. A pre-sale is worth ten interviews, because
   people say yes in interviews for free.
4. **Run it long enough and wide enough to fail.** A test that cannot produce a
   no is a marketing exercise.
5. **Read it against the bar, not against hope.** Record both.
6. **Close honestly.** `failed` and `inconclusive` are the results that save
   money, and a studio that never records one is validating nothing and building
   anyway.

## Output

A `validations` row with the bar it was judged against and the actual result. A
failed validation is a successful use of this skill.