Marketplace Conversion Audit

Most funnel pages lose people for ordinary reasons: no clear next step, a hidden price, no reviews or refund terms, a slow first response, a login wall, broken link previews. This skill finds those in minutes and tells you which to fix first.

A zero-dependency Node.js script fetches each page you list (GET and HEAD only) and reads the HTML: title, meta description, H1, viewport, canonical and Open Graph tags, JSON-LD such as Product, Offer and ratings, calls to action, visible price, trust signals, image alt text, form length, response time, HTML size and a sample of internal links. It scores every page against a weighted rubric for its page type, gives buyer and seller (or your own) segment scores, ranks the top 10 blockers with a concrete fix for each, and compares with the previous run. Reports come as Markdown and JSON, with exit codes for scheduled checks.

Your AI then reviews the copy of each page (offer clarity, objections answered, friction) and combines both halves into a short prioritised action list.

Inside: the script with --help, --dry-run and an offline --self-test, the rubric and config references, an AgentLoomX example config with buyer and seller funnels, store and SaaS templates, a review worksheet, and a real audit of agentloomx.com with a full walk-through. Old 1.x configs still run.

It sees public HTML only. It has no analytics, runs no JavaScript and does not log in, so its scores are coverage signals, not conversion rates. Connect real analytics to measure actual rates.