Skip to content
Back to skills

Ux Bar

ASecurity

UX quality bar — Krug's three laws of usability, the eight DevEx first principles, time-to-hello-world tiers, the Goodwill Reservoir, the Pit of Success.

  • 11 stars
  • 0 votes
  • 0 copies
  • 4 views
  • Added May 27, 2026
data-aigoreactawsbackend

Works with

  • cli

Security analysis

A100/100

Scanned May 27, 2026

npx -y skills add 0xabrar/gstack-distilled --skill ux-bar --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Ux Bar?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Ux Bar
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/0xabrar-ux-bar/badge)](https://www.skillsdirectory.com/skills/0xabrar-ux-bar)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: ux-bar
author: 0xabrar
description: UX quality bar — Krug's three laws of usability, the eight DevEx first principles, time-to-hello-world tiers, the Goodwill Reservoir, the Pit of Success.
---

# UX bar

What "good" feels like for users. Use when reviewing flows, onboarding, error messages, and dev experience.

Source: gstack `design-review/SKILL.md`, `devex-review/SKILL.md`, `qa/SKILL.md`.

## Krug's three laws of usability

1. **Don't make me think** — self-evident > self-explanatory > requires explanation
2. **Clicks don't matter, thinking does** — three mindless clicks beat one puzzle
3. **Omit, then omit again** — happy talk and instructions must die

Supporting doctrine:
- *"Users scan, they don't read… designing billboards going by at 60 mph."*
- *"Users satisfice — pick first reasonable, not best."*
- *"Clarity trumps consistency."*

## The Goodwill Reservoir

Every user starts with a finite reservoir of goodwill. Every friction point depletes it. When it hits zero, they leave.

Friction points that drain it:
- Error messages without fixes
- Required fields with no explanation
- "Click here to continue" buttons that lead to dead ends
- Slow loads
- Forms that lose state on error

Goodwill restorers:
- Errors that include the fix
- Defaults that just work
- Speed
- Surprise-and-delight moments

## Pit of Success

> "Make the right thing easy, the wrong thing hard."

If users keep doing the wrong thing, the design is wrong, not the users.

## Eight DevEx first principles

For tools where your users are themselves builders:

1. **Zero friction at T0** — first-run should require nothing the user doesn't already have
2. **Decide for me, let me override** — opinionated defaults, escape hatches everywhere
3. **Every error = problem + cause + fix** — never just "something went wrong"
4. **Speed is a feature** — latency in dev tools compounds
5. **Show, don't explain** — examples beat docs
6. **Composability over completeness** — a sharp tool you combine beats a Swiss Army knife
7. **Honest about state** — if the tool is in a weird mode, say so
8. **Chef-for-chefs bar** — *"your users build products for a living, the bar is higher"*

## Time-to-hello-world (TTHW) tiers

| Time | Tier | What it means |
|---|---|---|
| <2 min | **Champion** | Best-in-class. Users will recommend. |
| 2-5 min | **Competitive** | Acceptable. Most successful tools live here. |
| 5-10 min | **Needs Work** | Friction is hurting conversion. |
| **>10 min** | **Red Flag** | 50-70% abandonment. Fix this before anything else. |

If you can't measure TTHW for your product, you're flying blind on conversion.

## The boomerang (DevEx review)

At plan stage, score predicted N/10 on each UX dimension.

After ship, do a live audit and score actual.

**Flag any dimension where `live < plan - 2`.**

This forces honesty about the gap between what you intended and what you shipped.

## First Impression rule

When designing for review:

> "A designer doesn't hedge, they react. The first impression is gut-reaction first."

Don't ask reviewers what they think — watch them load the page and listen to the first thing they say.

## Trunk test

For any nav: if a user lands on a sub-page from search, can they figure out where they are and where to go in 2 seconds?

If not, the nav is broken.

## Errors are for AI agents now

> "Every error message must be actionable. The agent should be able to read the error and know what to do next without human intervention."

In an AI-first world, error messages aren't just for humans — they're inputs to a fix loop. "Something went wrong" wastes both human and agent time.

## Reproduce or it didn't happen (QA)

> "Repro is everything. Every issue needs a screenshot. No exceptions."

> "Depth over breadth — 5-10 well-documented issues > 20 vague descriptions."

> "Test like a user. Never read source code."

> "Never refuse to use the browser." — even backend-only diffs ship through the browser

## Three QA tiers

| Tier | Time | Scope |
|---|---|---|
| **Quick** | ~30s | Smoke test, top 5 nav |
| **Full** | 5-15 min | 5-10 well-evidenced issues |
| **Exhaustive** | Open-ended | Cosmetic + edge cases |

Triage tier dictates fix scope:
- Quick fixes critical + high
- Standard adds medium
- Exhaustive does cosmetic too

## Common UX anti-patterns

- "Click to continue" with no preview of what's next
- Errors without fixes
- Required fields with no inline validation
- Forms that clear on error
- Modals that block primary actions
- Onboarding that requires completing X steps before showing value
- Default opt-in for things that should be opt-out (and vice versa)

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.