Website Copy Analysis & Optimization Skill
Fetches and scores any web page's copy across five dimensions, then generates specific before/after rewrites using proven frameworks like PAS and AIDA.
The skill only reads a target URL via WebFetch and writes a local Markdown report; no credential access, destructive defaults, or covert exfiltration; narrow privileges and reasonable transparency. Deductions: no declared fetch boundaries or rate limits, no user-confirmation or rollback disclosures; publisher unverified — full marks not justified.
SKILL.md is internally consistent with a clear fetch→analyze→score→generate flow and complete output templates. Deductions: failure handling for WebFetch errors, JS-rendered pages, or non-English content is undefined; no tests or key-path reproduction evidence, so the static anchor cap of 10 applies and abnormal-input behavior is thin — scored at the lower end.
Trigger condition (/market copy <url>) is explicit, page-type mapping is clear, and cross-skill integration is documented. Deductions: non-fit ranges (login-walled pages, non-marketing pages) undeclared, no Chinese/multilingual support, and dependence on WebFetch reachability may impair use from mainland-China networks without disclosure.
Well-layered documentation consistent with the README architecture; MIT license and uninstall script present. Deductions: no versioning/changelog in the skill file, no known-limitations disclosure, no stated maintenance owner or update path, and hidden assumptions (WebFetch availability, date source) unexplained.
Outputs (scores, before/after pairs, swipe file, COPY-SUGGESTIONS.md) are well specified and directly usable, with real marginal value over manual analysis. Deductions: 'proven framework' claims have no third-party execution evidence, scoring rubrics are uncalibrated subjectivity, and the static anchor cap of 7 applies with limited completeness evidence — mid-low score.
Only author claims and README demo terminal output (illustrative, not executed proof). Deductions: no CI evidence, no committed tests, no cross-source corroboration; key claims (scoring methodology, framework efficacy) are not independently reproducible — author-claims level, low score.
- The skill depends on WebFetch to crawl external websites; fetching overseas sites from mainland-China networks may fail, and this limitation is not disclosed.
- No tests or execution evidence exist; scoring rubrics are subjective and output quality is unverified — review results manually before use.
- The skill file has no versioning or changelog, the publisher is unverified, and long-term maintenance is unclear.
- Do not use on login-walled pages, sensitive data, or sites you do not own or lack rights to analyze.
What does this skill do, and when should you use it?
This is the copywriting sub-skill of the AI Marketing Suite for Claude Code, triggered by the /market copy <url> command. It fetches a target web page, detects its page type (homepage, landing page, pricing page, etc.), analyzes the existing voice and value proposition, and scores the copy across clarity, persuasion, specificity, emotion, and action (out of 50, scaled to 100). It then generates 5-10 alternative headlines using PAS, AIDA, Before-After-Bridge, and 4U frameworks, plus at least five before/after rewrite pairs covering headline, subheadline, CTA, body copy, and meta description. Results appear as a terminal summary, with the full report written to COPY-SUGGESTIONS.md.
1) Uses WebFetch to retrieve the target URL and extract headlines, subheadlines, section headers, body copy, CTA button text, navigation, footer, meta tags, and social proof; 2) detects page type (homepage, landing, pricing, about, product, feature, blog, contact/demo) and applies type-specific copy priorities; 3) analyzes brand voice across formality, emotion, complexity, humor, and authority; 4) scores the headline and full page copy on 0-10 rubrics; 5) maps a value proposition canvas and flags gaps; 6) generates 5-10 framework-based headline alternatives, CTA optimization advice (button text, placement, color psychology), and at least 5 before/after pairs; 7) builds a swipe file of ranked headline, subheadline, CTA, and meta alternatives; 8) prints a condensed terminal summary and writes the full structured report to COPY-SUGGESTIONS.md. It references BRAND-VOICE.md, MARKETING-AUDIT.md, or COMPETITOR-REPORT.md when present.
- An agency builder auditing a prospect's landing page before a sales call to produce quantified copy fixes for a pitch
- A solopreneur diagnosing their own homepage for vague headlines and weak CTAs, with concrete rewrites
- An e-commerce seller optimizing a product page with benefit-driven description and meta description rewrites
- A SaaS team improving pricing page copy, including plan naming, anchoring, and objection-handling FAQ advice
- A marketing consultant generating client-ready structured copy diagnostics (COPY-SUGGESTIONS.md) across multiple client sites
What are this skill's strengths and limitations?
- Structured, transparent methodology: explicit scoring rubrics, page-type differentiation, and proven frameworks (PAS, AIDA, BAB, 4U)
- Every recommendation requires a concrete before/after example with rationale, not vague commentary
- Voice-aware: analyzes existing tone and keeps generated copy consistent with it
- Integrates with other suite artifacts (BRAND-VOICE.md, MARKETING-AUDIT.md, COMPETITOR-REPORT.md) for cumulative context
- Produces a client-ready Markdown report with impact-ranked implementation priorities
- Depends on WebFetch, so it is limited to publicly fetchable pages; handling of login-gated or heavily JS-rendered pages is undocumented
- No test suite; rewrite quality rests entirely on model capability, and example claims (e.g. the sample "40%" figure) are illustrative format, not real outcomes
- Explicitly triggered only via the Claude Code /market copy slash command; invocation on other platforms is undocumented
- Scores are heuristic model judgments, not validated against real conversion data
- No offline or local-file analysis mode — a URL is required
How do you install this skill?
The skill is one of 14 sub-skills in the ai-marketing-claude repository and installs with the suite:
- One-command: curl -fsSL https://raw.githubusercontent.com/zubair-trabzada/ai-marketing-claude/main/install.sh | bash
- Or manual: git clone https://github.com/zubair-trabzada/ai-marketing-claude.git && cd ai-marketing-claude && ./install.sh
The skill file lives at skills/market-copy/SKILL.md. Standalone installation of this sub-skill is not documented.
How do you use this skill?
In Claude Code, run /market copy <url>, e.g.: /market copy https://calendly.com. The skill fetches the page, scores it, and generates suggestions — a condensed summary appears in the terminal and the full report is saved to COPY-SUGGESTIONS.md in the working directory. Optional follow-ups: /market landing for a landing-page deep dive, /market brand for voice guidelines.