Improve a Website
Diagnose, fix, and prove website improvements
Takes you from a live website that underperforms to a prioritized, evidence-backed backlog of conversion, usability, message, and speed fixes - each shipped as a testable experiment. It orchestrates its eight skills phase by phase, asking you the decision questions at every step and recording every result in your project's docs/ folder so the journey resumes across sessions.
npx skills add wondelai/skills/improve-website --global - 01
Why don't visitors convert, and which fix moves the most money? gate
CRO Methodology → docs/METRICS.md + docs/WEBSITE.md + docs/EXPERIMENTS.md - 02
Where do visitors stumble before they weigh the offer?
UX Heuristics → docs/DESIGN.md + docs/EXPERIMENTS.md - 03
Does the page look as clear as it reads?
Refactoring UI → docs/DESIGN.md + docs/EXPERIMENTS.md - 04
Can visitors comfortably read the copy?
Web Typography → docs/DESIGN.md + docs/EXPERIMENTS.md - 05
Does a stranger grasp the offer in five seconds?
StoryBrand Messaging → docs/POSITIONING.md + docs/EXPERIMENTS.md - 06
Does the page arrive before patience runs out?
High Perf Browser → docs/METRICS.md + docs/WEBSITE.md + docs/EXPERIMENTS.md - 07
Is the idea memorable enough to survive to the decision?
Made to Stick → docs/POSITIONING.md + docs/EXPERIMENTS.md - 08
Do the visitors who act ever get stuck?
Design of Everyday Things → docs/DESIGN.md + docs/EXPERIMENTS.md
Each phase asks the question, then writes the answer before the next one starts.
The long read behind the phases: How to Improve an Existing Website with AI Skills
02 What it is
What is the Improve a Website journey?
Traffic arrives, conversions do not. This journey turns that gap into a ranked, evidence-backed queue of fixes across conversion, usability, hierarchy, typography, message and speed. Eight phases, each posing one decision question and writing its findings to docs/, so a redesign never becomes a guess you cannot revert.
03 In practice
Fix a Website That Doesn't Convert
Traffic is fine; signups aren't.You want evidence-driven fixes, not a redesign lottery.
Our website's conversion is too low. Run the improve-website journey: diagnose why visitors leave, then fix usability, hierarchy, typography, message, and speed.
Use improve-website skill.
Written to your repo
- docs/METRICS.md
- docs/WEBSITE.md
- docs/EXPERIMENTS.md
- docs/DESIGN.md
- docs/POSITIONING.md
- docs/IMPROVE-WEBSITE-PLAN.md journey tracker
Stop anytime. The tracker records which phases are done and which decision is next, so a new session opens at the unanswered question.
04 Prompts
Say it like this.
Name the journey at the end of a prompt and it takes over from phase one. Say it again in a later session and it resumes at the unanswered question.
Our landing page isn't converting — diagnose it and fix it phase by phase. using improve-website skill
Audit and improve our site end to end, evidence first. using improve-website skill
Resume the website improvement journey. using improve-website skill
05 Fit
One prompt runs it.
Every phase asks before it acts, and every decision lands in your project's docs/ folder as a living document.
Reach for it when
Find out why visitors don't convert before redesigning anything
Rank usability problems by severity so the worst ones get fixed first
Fix visual hierarchy, typography, and page speed systematically
Sharpen the message so a first-time visitor gets it in five seconds
Turn every change into an experiment you can prove or revert
Reach for something else when
You have no site yet. Create a Website builds one message-first instead of diagnosing an existing one.
Conversion is already healthy and you want more from it. Grow a Website layers capture, offer and shareability on top.
You only need a page-speed pass. High Perf Browser handles that alone, without seven other audits.
The interface is a product rather than a marketing page. Improve an App covers in-app flows and copy.
07 Questions
Before you install.
How do I install the Improve a Website journey?
Run npx skills add wondelai/skills/improve-website --global. It takes about 30 seconds and needs no account. The Improve a Website skill then works in Claude, Claude Code, Claude Cowork, Codex, Cursor, OpenClaw and Hermes Agent — anything that reads the open agentskills.io format — and your agent loads it on its own when a task calls for it. It is free and MIT-licensed, and the source is at https://github.com/wondelai/skills.
Which skills does the Improve a Website journey run, and what does it produce?
It orchestrates 8 skills across 8 phases — CRO Methodology, UX Heuristics, Refactoring UI, Web Typography, StoryBrand Messaging, High Perf Browser, Made to Stick, Design of Everyday Things — asking you one decision question per phase. Each answer is written to your project's docs/ folder (docs/METRICS.md, docs/WEBSITE.md, docs/EXPERIMENTS.md, docs/DESIGN.md, docs/POSITIONING.md, docs/IMPROVE-WEBSITE-PLAN.md), so the journey survives across sessions and you keep the reasoning, not just the output.
Why does conversion diagnosis come first and gate everything else?
Because every phase after it costs time, and without the diagnosis you cannot say which is worth running. The CRO phase maps the funnel, finds where the money leaks and ranks the candidates in docs/METRICS.md and docs/EXPERIMENTS.md. If the drop-off is in checkout, the typography phase can wait. The gate exists to stop the default failure mode: redesigning the page that was already fine.
Every phase writes to docs/EXPERIMENTS.md. Do I have to A/B test everything?
No — the file records the change, the reason and what you expect it to move, whether or not you run a formal test. Some fixes are obvious enough to ship straight: a broken form label, a four-second load. Others deserve a real experiment. What the journey refuses is an untraceable change — if conversion moves next month, docs/EXPERIMENTS.md is what tells you which of the eight phases moved it.
How does this differ from Grow a Website?
Improve fixes what is broken; Grow adds what is missing. This journey ranks defects — usability friction, weak hierarchy, unreadable type, a message nobody parses in five seconds, a slow page — and turns each into a revertible change. Grow a Website assumes those basics hold and builds lead capture, proof, a stronger offer and shareability on top. Run this one first if visitors are stumbling rather than merely unpersuaded.
Can I stop after the first two phases?
Yes, and many should. CRO and UX Heuristics usually account for the largest single movement, and docs/IMPROVE-WEBSITE-PLAN.md will hold the remaining six open indefinitely. Come back once those fixes have shipped and you have fresh numbers — the later phases on message, speed and affordances are more useful when the obvious leaks are already sealed.
Eight skills, one deliberate order.
Run the whole journey with one prompt.
npx skills add wondelai/skills/improve-website --global MIT licensed · Claude, Claude Code, Codex, Cursor, OpenClaw, Hermes · No account