Ship the change that gets one more visitor to yes.
Audits a landing page and returns prioritized conversion fixes across hero, proof, friction, and CTA.
Inside: The hero, proof, friction, and CTA checks, ranked by effort vs impact, with lift direction.
Install it in one line, or paste it in.
~/.claude/skills/ and runs automatically when it is relevant.Connect your context. Set it to your motion.
it sees bounce, scroll depth, and where visitors drop, so the fixes target the section that is actually leaking.
it can check the page against your other converting pages and reuse what already works.
This was built for a B2B signup or demo page. Set these to your stack:
Audit for one GOAL_ACTION per page. A page asking visitors to do three things converts on none. If the page has two asks, the skill flags it.
| Set this | What it is | Example |
|---|---|---|
| GOAL_ACTION | the one thing a visitor should do | start trialbook demosubscribe |
| AUDIENCE | who the page is for | foundersRevOpsIT buyers |
| PROOF assets | evidence you can show | logosquotesnumbersratings |
| ANALYTICS | your web-analytics connector | a web-analytics tool |
| CMS | where the page is edited | a CMS |
| ABOVE_FOLD | what shows without scrolling | headlinesubheadCTAone proof |
Everything the skill does, in full.
Reads a landing page and returns a ranked list of conversion fixes, grouped by the four things that move a page: the hero, the proof, the friction, and the CTA. Each fix says what to change, why it matters, and which way conversion should move if you make it. It is a worklist you can hand to whoever owns the page, not a vague "make it better."
- 1Hero (the first screen)
Checks the above-the-fold in one glance: does the headline say what this is and who it is for, is the value clear in five seconds, is the CTA visible without scrolling, and is there one proof element early. A hero that makes the visitor work is the most expensive leak on the page.
- 2Proof
Checks that claims are backed where the visitor doubts them: a specific outcome near the value claim, recognizable logos or quotes near the CTA, and real specifics over adjectives. "Trusted by teams" is not proof. A named result is. Missing proof is a named fix.
- 3Friction
Finds what makes saying yes harder than it needs to be: too many form fields, an unclear next step, a wall of copy before the point, a CTA that hides, competing links that lead visitors away. Each friction point is named with the fix.
- 4CTA
Checks the ask itself: is there one primary action, is the button text specific about what happens next ("Start free trial", not "Submit"), does it repeat at natural decision points, and does it match the GOAL_ACTION. Vague or stacked CTAs are flagged.
- 5Prioritized fix list
Every finding is scored effort against impact and sorted. Cheap high-impact fixes rise to the top, each tagged with the direction conversion should move. You ship the wins first.
- No claimed conversion rate without a connected analytics source. Unknown stays unknown.
- Lift is always a direction (up, likely up), never a fabricated percentage.
- Every fix names the specific element to change, never "improve the page."
- Illustrative bounce or conversion numbers are marked as examples, never presented as your real data.
LANDING PAGE CRO · page: /trial · goal: start free trial
Est. leak: hero + form (connect analytics to confirm)
Top fixes (effort -> impact, lift direction):
1. LOW -> HIGH ^ Headline says what you do, not who it is for.
Add the audience: "for RevOps teams."
2. LOW -> HIGH ^ CTA reads "Submit." Change to "Start free trial."
3. MED -> HIGH ^ Form asks for 7 fields. Cut to email + name.
Ask the rest after signup.
4. LOW -> MED ^ No proof above the fold. Move one customer quote up.
5. LOW -> MED ^ 3 nav links compete with the CTA. Remove them on
this page. One way out.
Next: connect a web-analytics tool to confirm which section leaks most.
Where an operator takes this next.
The audit is step one. Here is where an operator takes it once the manual version proves out.
You built the worklist once, now it keeps itself current with the page.
Wire a scheduled Claude task to pull the live page weekly and diff the fix list against what's shipped, so nothing regresses silently.
Connect Amplitude or a web-analytics tool so the ranked list targets the section actually losing visitors, not just the section that looks weak.
Push the #1 ranked fix into Linear or your CMS's task queue the moment the audit runs, so it doesn't sit in a doc nobody opens.
One skill is the on-ramp.
A single skill does one job. Chained into a playbook, or run as a full build, it becomes a system. Here is where this one plugs in.