# Validation Designer

> Design the smallest useful test for an idea, plan, hypothesis, or solution. Use when the next step should reduce uncertainty with evidence instead of more debate, and when a proposal needs a concrete validation path before larger execution.

- Skill: `alubiama/validation-designer` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add alubiama/validation-designer`
- Raw SKILL.md: https://api.skillmd.com/api/skills/alubiama/validation-designer/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: Alubiama (https://skillmd.com/u/alubiama)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/alubiama/validation-designer

---


# Validation Designer

## Goal

Turn uncertainty into the smallest useful validation step.

This skill does not judge whether the idea is good overall. It designs how to learn quickly enough to change the next decision.

## When To Use

- after `evidence-gap-detector`
- before committing to implementation
- when a proposal depends on one or two key assumptions
- when the team needs a test, not more argument

## Scope Boundaries

In scope:

- define the hypothesis to test
- choose the smallest validation method
- define success and failure signals
- recommend the next concrete check

Out of scope by default:

- full experiment programs
- proving everything at once
- broad product strategy

## Escalation Conditions

Pause and keep the design narrow when:

- the decision to inform is still unclear
- several uncertainties compete and none is clearly primary
- the proposed test is larger than the decision requires
- no result from the test would actually change the next move

If those conditions persist, route back to `evidence-gap-detector`, `quality-gate`, or `intent-clarifier` before designing a bigger validation path.

## Validation Workflow

1. State the decision that validation should inform.
2. Identify the highest-impact uncertainty.
3. Convert it into a testable hypothesis.
4. Choose the lightest valid method:
   - interview
   - prototype test
   - smoke test
   - instrumentation check
   - A/B or quasi-experiment
5. Define the minimum signal needed.
6. Define an `if-then` decision rule:
   - if signal appears, do X
   - if signal fails, do Y
7. Name the stop condition and next action.

## Output Contract

Always return:

1. `decision_to_inform`
2. `key_uncertainty`
3. `test_hypothesis`
4. `validation_method`
5. `success_signal`
6. `failure_signal`
7. `if_then_rule`
8. `effort_level`
9. `recommended_followup`
10. `next_test`

## Guardrails

- Do not design a bigger test than the decision requires.
- Do not confuse data collection with validation.
- Prefer reversible, fast checks first.
- If the uncertainty is not decision-relevant, say so.
- Tie the test to one primary uncertainty whenever possible.
- If the best next move is clarification rather than validation, say so.
- Make the result actionable by tying it to an explicit next decision.

## Anti-Patterns

- vanity experiments
- testing five assumptions at once
- no success threshold
- no decision tied to the result