Imported from Emmraan/agent-skills (skills/marketing-growth/improve-website/SKILL.md). Install upstream with npx skills add Emmraan/agent-skills --skill improve-website. Copyright stays with the author (MIT).

Improve a Website

An interactive, resumable journey that turns a live-but-underperforming website into a ranked backlog of defensible fixes. Across eight phases it diagnoses why visitors don't convert, then fixes usability, hierarchy, typography, message, speed, memory, and errors — asking you at each decision point and recording results in docs/. Each change is tied to a reason; the big ones ship as tests.

Core Principle

Diagnose, fix, prove: evidence from real visitors decides what changes, and every fix ships as a testable hypothesis. Order is deliberate — research first, then usability, look, message, speed, memory, and error-tolerance, because each phase's output feeds the next.

This skill sequences the phases, asks the decision questions, and records what you decide. The constituent skills carry the method — invoke them rather than improvising their frameworks.

Journey Map

Operating Rules

1. Resume first. Before anything else, read docs/IMPROVE-WEBSITE-PLAN.md and every artifact in the Journey Map. If the tracker exists, summarize the journey state in 3-5 lines and ask which phase to enter. Done when the user has confirmed an entry point. A journey with a tracker is resumed, never restarted.
2. Intake on first run only. No tracker: run the Intake below, then create docs/IMPROVE-WEBSITE-PLAN.md with every phase statused pending | in-progress | awaiting-evidence | done | deferred: reason | skipped: reason. Done when the tracker exists and the user has confirmed the phase plan.
3. Phase entry. Announce: what the phase does, the decision it forces, the artifact it produces, rough effort. Offer proceed / skip / defer — phases marked GATE may be deferred, never skipped. Mark the phase in-progress on proceed. Done when the user chose.
4. Skill invocation and fallback. Invoke the phase's skill by its slug. If it is not available, offer: npx skills add wondelai/skills/<slug> --global. If the user declines, run the phase from its Brief — the minimum viable method. State which mode you are in.
5. In-phase decisions. Ask every question under "Decide with the user" — with concrete options and your recommendation. Record the choice in the tracker's Key Decisions. A decision made silently is a defect.
6. Phase exit. Present the draft artifact content for sign-off before writing. On approval: write or extend the docs/ files, update the tracker (status, Key Decisions, Next Actions). Done when the files are written and the phase row shows done.
7. Artifact discipline. Read before writing; create a file only if missing, otherwise extend — add or update your sections, preserve everyone else's. Files are UPPERCASE in docs/. Every recommendation lands as a checkbox or a table row with owner and priority. See references/artifact-templates.md when creating a docs/ file for the first time — create it from the full skeleton (all section headings), then fill the sections your phase names.
8. Diagnose before redesigning. No visual or copy change ships before the Phase 1 diagnosis produces evidence for it, and every change lands in EXPERIMENTS.md with a pre-committed metric. A fix with no Phase 1 finding behind it goes back to Phase 1; a bold change with no test attached stays in the backlog until it has one.

Intake

Ask these before creating the tracker:

1. What page or flow are we improving, and what is the ONE action it should drive? (Scopes every phase; gates Phase 1 — a page with three CTAs has no goal.)
2. What is the conversion problem in your words, and what evidence do you have — analytics, bounce rate, drop-off? (Feeds the Phase 1 funnel diagnosis.)
3. Do you have real visitor input — exit surveys, reviews, support tickets, recordings — or can you get it? (Phase 1's counter-objections must use the customer's own words.)
4. Roughly how much traffic per week? (Decides whether A/B tests can reach significance; low traffic leans on the qualitative and heuristic phases.)
5. Which complaint do you hear most — "looks unprofessional," "nobody understands what we do," "it's slow," "the text is hard to read"? (Routes which fix phase to front-load, per the guide FAQ.)
6. Do you have existing positioning or brand docs (POSITIONING.md)? (Phase 5 and 7 build on them.)
7. Can you paste real assets — URLs, live copy, CSS, analytics screenshots? (The skills are only as good as the evidence you feed them.)

Skip heuristics: skip Phase 4 if body text already meets 16px / ~66ch / 1.5-1.7 line height; skip Phase 6 if Core Web Vitals are already green (but triage a fast LCP fix first if the page barely renders); skip Phase 5 or 7 if messaging is already validated and clear. Never skip Phase 1 — it aims every other phase.

Then create docs/IMPROVE-WEBSITE-PLAN.md from the template and confirm the plan. Done when docs/IMPROVE-WEBSITE-PLAN.md exists with every phase statused and the user has confirmed the plan.

Phases

Phase 1 — Diagnose why visitors don't convert (cro-methodology) — GATE

Purpose: Find where and why the page loses visitors before any pixel changes. Phases 2-8 may only fix problems traceable to a Phase 1 finding.

Brief (fallback): Don't guess — discover. Map the funnel for blocked arteries (underperforming high-traffic stages) and missing links (stages that don't exist). Name the ONE action each page drives. Build the O/CO table for the Big 5 (Trust, Price, Fit, Timing, Effort) in the customer's words, each counter at friction. Rank by ICE.

Invoke: cro-methodology with the target page/flow, the ONE action, and real visitor input. Ask for a funnel map, an O/CO table, missing persuasion assets, and an ICE-ranked hypothesis list.

Decide with the user: (1) Confirm the ONE action per page; cut competing CTAs. (2) Attack the highest-money blocked artery or a missing link first? (3) Low traffic: accept qualitative + heuristic evidence.

Artifact: Extend docs/METRICS.md ## Funnel and ## Stage & One Metric That Matters; docs/WEBSITE.md ## Conversion Elements and ## Audit Findings (issue | severity 0-4 | fix | status); docs/EXPERIMENTS.md ## Experiment Backlog (ICE-ranked). Update the tracker.

Done when: the funnel is mapped, the ONE action set per page, every researched objection has a counter, the backlog is ICE-ranked, and Phase 1 shows done — only then are Phases 2-8 unlocked.

Phase 2 — Remove the usability friction (ux-heuristics)

Purpose: Find where visitors stumble — the friction that makes them quit before they weigh the offer.

Brief (fallback): Don't make me think. Users scan, satisfice, and muddle through. Run a heuristic evaluation against Nielsen's 10 heuristics; rate each issue 0-4 by frequency, impact, and persistence so catastrophes outrank cosmetics. Apply the Trunk Test to interior pages (what site, page, options, where am I, where's search?). Cut half the words, then half again; kill mystery-meat icon navigation.

Invoke: ux-heuristics with the page/flow and the Phase 1 findings. Ask for a severity-rated evaluation, a Trunk Test result per key page, and rewritten error and label copy.

Decide with the user: Which severity-4 and -3 issues get fixed now versus backlogged? Confirm fixes run by severity, not by ease.

Artifact: Extend docs/DESIGN.md ## UX Audit Findings (issue | heuristic | severity 0-4 | fix | status); append each fix to docs/EXPERIMENTS.md ## Experiment Backlog with ICE. Update the tracker.

Done when: every issue carries a 0-4 severity, the Trunk Test passes on each key page or is logged, and severity-4 issues have an owner.

Phase 3 — Fix the visual hierarchy (refactoring-ui)

Purpose: Make the page look as clear as it reads — one obvious focal point, not everything competing.

Brief (fallback): Great UI is systems, not talent. Design in grayscale first: establish hierarchy with size, weight, and contrast before any color — combine levers, don't multiply (all three only for the hero element). Enforce a spacing scale (4/8/16/24/32/48/64) where gaps between groups exceed gaps within them. Constrain text to 45-75 chars. Add color last: 5-9 shades per hue, grays tinted, not #000.

Invoke: refactoring-ui with the page, the Phase 2 findings, and the CSS framework. Ask for a grayscale hierarchy pass, spacing-scale corrections, a systematic palette, and exact class/CSS changes.

Decide with the user: Fix within the current design system or introduce new tokens? Confirm color work waits until the grayscale layout reads.

Artifact: Extend docs/DESIGN.md ## Tokens (spacing scale · palette shades · shadows) and ## Components (component | decision | status); append fixes to docs/EXPERIMENTS.md ## Experiment Backlog. Update the tracker.

Done when: the layout reads in grayscale, the spacing scale is applied, one primary CTA dominates, and the token decisions are recorded.

Phase 4 — Make the words readable (web-typography)

Purpose: Fix the measurements that decide whether visitors can comfortably read the copy.

Brief (fallback): Size, line length, and line height matter more than the typeface. Fix three: body ≥16px (18px for reading-heavy pages); line length 45-75 chars (~66 sweet spot) via ch or max-width; line height 1.5-1.7 body, 1.1-1.25 headings. Build a modular scale; load fonts with swap, preload one, subset, <200KB.

Invoke: web-typography with the reading-heavy templates and the DESIGN.md type decisions. Ask for corrected CSS with a fluid clamp() scale and a font-loading plan.

Decide with the user: Keep the current typefaces or repair the pairing? Confirm the font-loading budget (<200KB) that Phase 6 will verify.

Artifact: Extend docs/DESIGN.md ## Typography (typefaces · scale · measure · line height · loading strategy); append fixes to docs/EXPERIMENTS.md ## Experiment Backlog. Update the tracker.

Done when: body size, measure, and line height meet targets on the key templates, the scale is defined, and the loading strategy is recorded.

Phase 5 — Sharpen the message (storybrand-messaging)

Purpose: Make a stranger understand what you do and why it matters in five seconds.

Brief (fallback): The customer is the hero; your brand is the guide (Yoda, not Luke). Run SB7: a Character wanting one thing, a Problem at three levels (external / internal / philosophical — most brands miss the internal), a Guide with empathy and authority, a Plan of 3-4 safe steps, one Direct plus one Transitional CTA, and Failure/Success stakes. Write the "We help [X] who struggle with [Y] to [Z]" one-liner.

Invoke: storybrand-messaging with the current copy, POSITIONING.md if it exists, and the internal problems from the Phase 1 objections. Ask for above-the-fold rewrites, a BrandScript, and one-liners.

Decide with the user: Which one-liner passes the repeat-after-one-hearing test? Confirm the internal problem the copy must name.

Artifact: Extend docs/POSITIONING.md ## Brand Script (StoryBrand), ## One-Liner, and ## Key Messages (surface | message | status); append copy tests to docs/EXPERIMENTS.md ## Experiment Backlog. Update the tracker.

Done when: the one-liner is chosen, above-the-fold copy names the internal problem with one Direct and one Transitional CTA, and the messages are logged by surface.

Phase 6 — Make the page fast (high-perf-browser)

Purpose: Ensure a slow load doesn't waste the clarity work — the page must arrive before patience runs out.

Brief (fallback): Latency, not bandwidth, is the bottleneck — cut round trips. Target Core Web Vitals: LCP <2.5s, INP <200ms, CLS <0.1, TTFB <800ms. Preload the LCP hero element; reserve space (width/height or aspect-ratio) to stop CLS; defer non-critical JS; inline critical CSS; cache immutable; on HTTP/2 undo sharding.

Invoke: high-perf-browser with the page URL and a performance trace or Lighthouse run if available. Ask for the LCP element, layout-shift sources, blocking resources, and a prioritized fix list.

Decide with the user: If the page barely renders, triage a fast LCP fix before the rest? Which fixes ship now versus need a test?

Artifact: Extend docs/METRICS.md ## Baselines & Targets (CWV metric | baseline | target | miss response) and docs/WEBSITE.md ## Audit Findings; append fixes to docs/EXPERIMENTS.md ## Experiment Backlog. Update the tracker.

Done when: CWV baselines and targets are recorded, every fix has an owner, and any HTTP/1.1 workaround still active on HTTP/2 is flagged.

Phase 7 — Make the core idea stick (made-to-stick)

Purpose: Make the offer memorable enough to survive the gap between the visit and the decision.

Brief (fallback): Beat the Curse of Knowledge. Score key messages on SUCCESs (Simple, Unexpected, Concrete, Credible, Emotional, Stories). Concrete: replace abstraction with sensory specifics ("order in 30 minutes, still hot"; "save 16 hours a month"). Simple: the Commander's Intent — the one thing visitors must remember. Make statistics human-scale (Sinatra Test); prefer specific over round figures.

Invoke: made-to-stick with the headline, value prop, and key features plus the Phase 5 one-liner. Ask for a SUCCESs score per message with concrete rewrites, the Commander's Intent, and a human-scale stat.

Decide with the user: What is the single Commander's Intent message? Which abstractions become which concrete specifics?

Artifact: Extend docs/POSITIONING.md ## Key Messages with SUCCESs scores and concrete rewrites; append copy tests to docs/EXPERIMENTS.md ## Experiment Backlog. Update the tracker.

Done when: each key message has a SUCCESs score and a concrete rewrite, the Commander's Intent is named, and the key statistic is human-scale.

Phase 8 — Design out the errors (design-everyday-things)

Purpose: Protect the visitors who decided to act — make the wrong action impossible, not just warned against.

Brief (fallback): There is no human error, only bad design. Bridge two gulfs: Execution ("how do I do this?") with clear signifiers (pressable buttons, editable fields) and constraints (date picker over free text, Submit disabled until valid); Evaluation ("what happened?") with feedback within 0.1s. Fix slips with undo and separated destructive actions; error messages say what went wrong and how to fix, no blame.

Invoke: design-everyday-things with the conversion-critical flows (signup, checkout, account setup). Ask for weak signifiers, where constraints make errors impossible, feedback gaps, and message rewrites.

Decide with the user: Where should a constraint replace an error message? Where should undo replace an are-you-sure confirmation?

Artifact: Extend docs/DESIGN.md ## UX Audit Findings (issue | heuristic = Norman gulf | severity 0-4 | fix | status); append fixes to docs/EXPERIMENTS.md ## Experiment Backlog. Update the tracker.

Done when: each conversion flow has signifier/constraint/feedback fixes logged, error messages meet the checklist, and destructive actions offer undo.

Optional Phases

Optional phases follow the same operating rules; insert where the Add-when condition first becomes true.

Common Mistakes

Completing the Journey

Exit checklist:

• Every phase in docs/IMPROVE-WEBSITE-PLAN.md is done or skipped: reason.
• WEBSITE.md, DESIGN.md, POSITIONING.md, and METRICS.md carry every fix as a table row with owner and priority.
• Each high-ICE change is a designed A/B test in EXPERIMENTS.md ## Experiment Cards with a pre-committed metric.
• The ONE action per key page is unambiguous and competing CTAs are removed.

Close the loop by re-invoking cro-methodology to design the bold tests for the high-ICE backlog: size the sample up front, run one full business cycle, require 95% confidence, never peek early — then promote them from ## Experiment Backlog to ## Experiment Cards, and close the tracker with Next Actions carried into the artifacts.

Forward routing: when the site converts and needs more qualified traffic, continue with grow-website. When the friction lives in the product, not the marketing site, continue with improve-app.