FunnelEnvy Skills

AI-powered positioning and competitive research skills for Claude Code. Built by FunnelEnvy.

Works standalone. Works better with FunnelEnvy's private data layer.

Skills

Quick Start

Clone the repo and work from inside it. Claude Code discovers skills automatically.

git clone https://github.com/FunnelEnvy/funnelenvy-skills.git
cd funnelenvy-skills

Run a skill:

/positioning-framework https://example.com
/positioning-update
/ga4-audit properties/123456789
/aa-audit --config path/to/config.json
/hypothesis-generator
/landing-page-generator example-co campaign-slug --stage all
/landing-page-generator example-co campaign-slug --stage brief
/voice-inference https://example.com
/voice-inference https://example.com --mode compare
/experiment-mockup 1

The --stage flag on landing-page-generator controls which phases run: brief, copy, design, qa, or all (default).

Research output goes to .claude/context/. Deliverables go to .claude/deliverables/.

Positioning Framework Depth Levels

The positioning-framework skill supports three depth levels. Other skills do not use the --depth flag.

Each depth builds on prior work. Running quick then standard then deep is incremental, not redundant.

/positioning-framework https://example.com --depth quick
/positioning-framework https://example.com --depth deep
/positioning-framework https://example.com --competitive-focus "Acme Corp"
/positioning-framework https://example.com --property properties/123456789

Output Files

Context (.claude/context/)

Deliverables (.claude/deliverables/)

How It Works

Skills build on each other. Each one reads from and writes to .claude/context/, creating a shared knowledge layer that downstream skills consume.

positioning-framework researches a company and its competitors, then produces structured context files with evidence-backed analysis. It runs autonomous web research across multiple source tiers (website, reviews, Reddit, SEC filings, job postings) depending on depth level. At standard and deep depth, render-default-deliverables runs automatically after it completes.

ga4-audit pulls 10-15 reports from a GA4 property via direct API (preferred) or analytics-mcp fallback. Classifies conversion events, groups pages by type, sizes opportunities, discovers element-level interactions (CTA clicks, link text, custom parameters), and detects failure modes. Produces performance-profile.md. Standalone -- works with or without positioning context, though it can optionally enrich its output with product-line mappings from company-identity.md if one exists. Requires GA4 credentials (see skills/ga4-audit/.env.example) or analytics-mcp as fallback.

aa-audit is the Adobe Analytics equivalent of ga4-audit. Runs a Python script against the AA 2.0 Reporting API and produces the same performance-profile.md schema, so all downstream skills (hypothesis-generator, render-default-deliverables) work identically regardless of analytics platform. Requires Python 3 with requests, Adobe Analytics API credentials (env vars), and a client config JSON file.

hypothesis-generator reads everything the other skills produced and generates a prioritized experiment roadmap. Without GA4 data, it works from positioning gaps alone (confidence capped at 4/5). With GA4 data, it unlocks 19 performance-driven triggers, calibrates ICE scores using real traffic numbers, adds baseline metrics and test feasibility estimates to each hypothesis, and routes infeasible experiments (insufficient traffic) to "What's Not Here" with alternative approaches.

voice-inference analyzes how a company communicates by extracting 12-15 pages across content types (homepage, product, blog, case studies, about) and building an evidence-backed voice profile. Scores tone dimensions, identifies vocabulary patterns and sentence architecture, catalogs 33+ categorized examples, and derives actionable voice rules. Two modes: observe (infer from content alone) and compare (compare inferred voice against customer-provided brand docs). Does not require positioning-framework to have been run first. Produces brand-voice.md.

experiment-mockup (in active development) takes a hypothesis from the experiment roadmap and builds a visual mockup showing the proposed change in the context of the real target page. In live mode (requires Chrome DevTools MCP), it injects the change into the user's browser, matches the site's design system using computed styles, and iterates on placement and styling in real time. In static mode (automatic fallback), it extracts page HTML and builds a standalone mockup file. Both modes produce a CRO placement rationale explaining why the element is positioned where it is, what visual hierarchy strategy it uses, and how the dev team should implement it.

positioning-update applies client feedback, stakeholder corrections, and new intelligence to existing context files. Paste an email, Slack thread, or meeting notes and it classifies each piece of information, shows you a structured change plan, and executes surgical edits after approval. No web research. Triggers deliverable re-render automatically.

render-default-deliverables converts context files into polished, shareable documents. No research, no analysis. Pure synthesis and formatting. Run it manually with /render-default-deliverables any time after editing context files.

Recommended Order

# 1. Build positioning context (who you are, your market, your competitors)
# Add --property to use GA4 data for page selection (optional)
/positioning-framework https://example.com --property properties/123456789

# 2. Apply client feedback to correct and enrich context (optional)
/positioning-update

# 3. Pull analytics data (what's actually happening on your site)
/ga4-audit properties/123456789

# 4. Generate data-informed experiment ideas
/hypothesis-generator

# 5. Analyze brand voice (standalone, works without positioning context)
/voice-inference https://example.com

# 6. Visualize specific experiment changes as mockups
/experiment-mockup 1

# 7. Re-render deliverables any time context changes
/render-default-deliverables

Each step is optional and independent, but they compound. Positioning context makes the analytics audit smarter (product-line page grouping). Analytics data makes hypothesis-generator smarter (traffic-calibrated scores, performance-driven triggers). Run what you have access to.

Prerequisites

Most skills are pure markdown with no external dependencies. These are the exceptions:

License

MIT