Landing Page CRO Analysis Skill
Runs a section-by-section conversion rate optimization teardown on any landing page, producing prioritized, actionable fixes.
Purely analytical prompt skill: no writes beyond generating LANDING-CRO.md, no credential handling, no destructive defaults, data flow largely transparent. Deducted for: undeclared permission boundaries around browser tools used to fetch pages, no confirmation/recovery disclosure, and unverified publisher identity.
Instructions are internally consistent; the 7-step framework maps cleanly to the output template with explicit scoring rubrics. Deducted for: many checklist items (load time, mobile behavior) require runtime data the model may not have, no abnormal-input handling, no tests, no defined failure feedback.
Trigger conditions are clear (URL plus CRO/landing feedback request) and page-type taxonomy is specific. Deducted for: no declared non-fit boundaries (inaccessible or non-English/Chinese pages), no Chinese-language support notes, and dependence on fetching overseas websites with no mainland-China reachability assessment.
Well-layered, readable docs with a complete output template; repository carries MIT license. Deducted for: no skill-level version, changelog, or known-limitation disclosure; unsourced benchmarks; unclear maintenance responsibility; install relies on remote curl|bash while the skill omits dependency notes.
Concrete, deliverable output format (prioritized fixes, A/B hypotheses, copy scoring) with plausible marginal value. Deducted for: static review cannot verify output quality; quantified revenue-impact claims (e.g., '15-30% click lift') are unsubstantiated; several audit dimensions exceed text-model capability, so outputs may need substantial human review.
Only author claims and a simulated demo transcript in README; no committed tests, no CI evidence covering this skill, no third-party reproduction. Deducted for: unsourced industry conversion and page-speed benchmarks throughout.
- Installer uses remote curl|bash execution; prefer cloning and reviewing before running
- Conversion-rate and page-speed benchmarks in the skill are unsourced; verify before client delivery
- Metrics like load time and mobile behavior exceed direct model observation; related findings may be inference, not measurement
- No declared handling for inaccessible or non-English pages; Chinese-site scenarios unvalidated
- Publisher is unverified; maintenance and update path are unclear
What does this skill do, and when should you use it?
This is one of the 15 marketing skills bundled in the zubair-trabzada/ai-marketing-claude repository, focused specifically on landing page Conversion Rate Optimization (CRO). It scores a page across 7 weighted sections (hero, value proposition, social proof, features/benefits, objection handling, CTA, footer), then layers on copy scoring, form, mobile, and page-speed impact audits. The output is a structured LANDING-CRO.md report with an overall score, per-section findings and fixes, A/B test hypotheses, and a prioritized fix list. It suits marketers and agencies who want an objective, deliverable assessment of their own or a client's landing page.
Given a landing page URL, it: 1) classifies the page type (lead capture, SaaS signup, e-commerce product, etc. across 8 types) and applies matching conversion benchmarks; 2) runs a 7-point CRO framework with per-section checklists and 1-10 scores (hero weighted 25%, value proposition 20%, social proof 15%, features 15%, objections 10%, CTA 10%, footer 5%); 3) scores copy on clarity, urgency, specificity, proof, and action orientation for a score out of 100; 4) audits form field count, labels, and error handling; 5) assesses mobile usability and load-time conversion impact (e.g., 3-5 second loads lose roughly 20% conversions); 6) generates A/B test hypotheses in 'if-then-because' format plus heat map interpretation guidance; 7) writes a LANDING-CRO.md report with quick-win/medium-term/strategic prioritized fixes and text-based before/after wireframes.
- A solopreneur whose new signup landing page converts below industry benchmarks wants ranked, effort-to-impact fixes.
- An agency needs a structured, client-deliverable CRO audit of a prospect's existing landing page before pitching optimization services.
- A growth team building a long-term A/B testing roadmap wants hypothesis-formatted experiment ideas (headlines, CTA copy, form field counts).
- A paid-ads manager wants to plug high-impact leaks in the hero, page speed, and mobile usability before scaling spend.
- A user who already ran the repo's /market audit wants a deeper, page-level conversion deep-dive on one specific landing page.
What are this skill's strengths and limitations?
- Very concrete methodology: each section has a checkable checklist, explicit scoring rubrics, and industry conversion benchmarks (e.g., 5-10% is good for lead capture), rather than vague advice.
- Structured, deliverable output: LANDING-CRO.md includes an overall score, per-section findings, a five-dimension copy score, A/B test hypotheses, and three-tier prioritized fixes.
- Recommendations must be tied to revenue impact and ranked by effort-to-impact ratio, avoiding 'change your button color' filler.
- Ships reference tables (page-type benchmarks, speed impact, form best practices) usable as standalone evaluation criteria.
- The skill itself does not orchestrate subagents; the '5 parallel subagents' belong to the repo's /market audit orchestrator and must not be attributed here.
- Analysis quality depends on the model's ability to actually fetch the page; the source does not explain how the URL is retrieved (browser tools may be needed) and points no dedicated fetch script at this skill.
- Benchmark figures (e.g., each extra form field cuts conversion ~7%) are embedded reference data without cited sources; verify before client-facing use.
- No test suite or scoring-consistency validation; repeated analyses of the same page may produce different scores.
- The heat map section is layout-based inference guidance, not analysis of real heat map data.
How do you install this skill?
Install the full skill collection: curl -fsSL https://raw.githubusercontent.com/zubair-trabzada/ai-marketing-claude/main/install.sh | bash, or git clone the repo and run ./install.sh (skills go to ~/.claude/skills/, subagents to ~/.claude/agents/). Uninstall with ./uninstall.sh. This skill lives at skills/market-landing/SKILL.md in the repo and can be copied manually. The source does not document whether installing this skill standalone is supported.
How do you use this skill?
In Claude Code, type /market landing <url> or /market cro <url>, or simply provide a landing page URL and ask for a CRO analysis. If /market audit was run previously, the skill incorporates those findings. Review the generated LANDING-CRO.md in the project root or output directory.