Skills directory
Discovery5-point gate · Any stack

Steal the bar, not the pixels

Tear down the three best competitors before designing anything.

Use this when

Designing from memory of a competitor

'They probably use X' stated as fact

Copying a layout without knowing why it works

The quality gate

  • Three best-in-class products torn down from screenshots, not memory
  • For each: what they observably do best (load choreography, states, copy, pricing framing)
  • The bar is written: 'ours must feel at least as X as Y's Z'
  • One deliberate divergence chosen — and why it wins for OUR user
  • Inferred internals are labeled inference, never presented as fact

Prove it: Show the written bar next to screenshots; a teammate can tell which product set each line.

Install or copy

Save the download as competitive-teardown/SKILL.md in your agent's supported skills directory, or copy the complete package below. Load it only when its description matches the work.

Portable SKILL.md
---
name: competitive-teardown
description: "Applies “Steal the bar, not the pixels” as a repeatable product-quality gate. Use during discovery work when tear down the three best competitors before designing anything."
---

# Steal the bar, not the pixels

Apply this skill during **Discovery** work. It supports Any stack.

Tear down the three best competitors before designing anything.

## Instructions

Before designing any significant surface, tear down the three best products with a similar job. Work from real screenshots and recordings, not memory. For each, extract what is observably excellent: first-load choreography, empty/loading/error states, copy tone, information density, pricing presentation, motion discipline. Write the explicit quality bar as sentences ('our onboarding must reach first value faster than X's'). Choose one deliberate divergence and defend it in one paragraph. Never dress inference as fact — 'their backend probably…' is labeled speculation or cut. The goal is their bar, not their pixels: copying layouts produces a worse version of someone else's product.

## Detect the problem

- Designing from memory of a competitor
- 'They probably use X' stated as fact
- Copying a layout without knowing why it works

## Hold the gate

Do not declare this phase complete until every item passes:

- [ ] Three best-in-class products torn down from screenshots, not memory
- [ ] For each: what they observably do best (load choreography, states, copy, pricing framing)
- [ ] The bar is written: 'ours must feel at least as X as Y's Z'
- [ ] One deliberate divergence chosen — and why it wins for OUR user
- [ ] Inferred internals are labeled inference, never presented as fact

## Verify the result

Show the written bar next to screenshots; a teammate can tell which product set each line.

More discovery skills

Want this quality gate applied to a real product decision?

Start a First Loop