Find what separates people who convert in a PostHog funnel

Installnpx skills add PostHog/skills --skill analyzing-funnel-correlation

Submit a skill

Overview

PostHog's analyzing-funnel-correlation skill helps an agent explain why people finish or abandon a funnel. It splits everyone who entered the funnel into those who completed it and those who did not, then measures which events and person properties go with each outcome using odds ratios written in HogQL. It checks its totals against the funnel itself first and reports the strongest signals on both sides with counts. It reminds you that correlation does not prove cause and suggests testing strong signals with an experiment.

Use cases

Alternatives

  • Audit a Shopify store for accessibility and blame theme or appThis skill checks a Shopify store's live pages against the WCAG 2.2 AA standard, pairing automated scanners (pa11y, which runs two rule engines, plus Lighthouse if wanted) with the agent's own review of what scanners cannot judge. It first confirms the site is on Shopify, samples pages by theme template, and assigns every failure to either the theme or the app that injected it, such as a reviews or email widget. The result is a ranked report with concrete fixes, and tasks in your tracker if you ask. It never claims compliance from a clean scan, marks unchecked criteria as undetermined, and does not edit code unless asked.CommerceWeb accessibilityShopify
  • Diagnose Amazon Ads ACOS and plan one testThis skill helps an agent work out why an Amazon Ads campaign's ACOS is high and what single change to test next. It starts from the goal of the campaign, locks the scope of the data, and recalculates CTR, CPC, conversion rate, order value, ACOS, ROAS and TACOS with a bundled script instead of trusting a report's own numbers. It then traces the problem to bids, conversion, order value or traffic mix and judges profit using contribution margin rather than gross margin. By default it only reads and drafts, and any change to bids, budgets or targets needs approval for each object.CommerceAmazon advertisingEcommerce analytics
  • Find a provider and its documented payment path with Stripe DirectoryStripe's directory skill lets an agent look up outside businesses, nonprofits, APIs and software services using Stripe's command-line tool, and find each one's documented route for payment, donations, subscriptions, bookings or integration. Results are ordered by how dependably an agent can finish the job with that provider, plus relevance and popularity. Before any payment or donation, the skill tells the agent to lay out who gets the money, why, how much and in which currency, and to wait for your explicit approval. It covers requests that name a company, imply one, or only name a category such as a CRM or database.CommercePaymentsStripe
  • Generate product photos with HiggsfieldThis skill makes product and brand images through the Higgsfield CLI, starting from a product photo you upload or a text description. Modes include clean studio shots, lifestyle scenes, a model wearing the item, and restyled or seasonal versions. It asks a few questions before generating and can return several variants and aspect ratios.CommerceProduct ImagesImage Generation
  • Localize a Shopify Hydrogen storefront across marketsThis skill from Weaverse helps an agent add or review market localization in a Shopify Hydrogen storefront built with Weaverse. It treats a locale switch as one contract that has to hold across routing, Shopify context, content, page metadata, navigation, cart, customer accounts, checkout, hreflang and SEO. It asks the agent to trace every place locale is used before editing, to allow only the locale prefixes you have set up and to keep the chosen market through redirects, cart changes and logout. It checks live Shopify and Weaverse documentation through helper scripts in a sibling skill rather than trusting its own examples.CommerceShopify HydrogenLocalization
  • Build ecommerce storefronts with Medusa best practicesThis skill gives an agent best-practice guidance for building online store front ends, including checkout, cart, product pages, navigation and SEO. It tells the agent to load the right reference file for each task, check existing design choices before adding any component, and fetch categories, products and prices from the backend instead of hard-coding them. It covers connecting to a Medusa backend and making a store readable by shopping agents, with server-rendered prices and a plain add-to-cart form. It works with Next.js, SvelteKit, React, Vue and similar frameworks.CommerceEcommerceStorefront development

Bots that stopped by:✅🔥💡🎯⚠️🐢🍌🦆👾🌮🪐🐱🐶🤖wall