FunnelEnvy
9 verified skills9 total stars
hypothesis-generator
When the user wants to generate experiment hypotheses from existing positioning context. Also use when the user mentions 'hypotheses,' 'experiment ideas,' 'test roadmap,' 'what should we test,' 'CRO opportunities,' 'A/B test plan,' or 'experiment backlog.' Reads L0 + L1 context files from .claude/context/, applies CRO reasoning patterns, and produces a prioritized, sequenced experiment plan in .claude/deliverables/. In KB mode (see KB Mode (Dual-Mode Output)), reads the scope's silver CRO artifacts from a bound knowledge base and writes a typed gold-experiment-roadmap artifact instead. No research, no web fetches. Analysis-grade synthesis using embedded CRO expertise.
development1
positioning-framework
When the user wants to build, audit, or update a positioning and messaging framework for a company or product. Also use when the user mentions 'positioning,' 'messaging framework,' 'competitive analysis,' 'competitive research,' 'battle cards,' 'competitive landscape,' 'value props,' 'persona messaging,' 'differentiation,' 'quick positioning,' 'positioning readout,' or wants to define how a company communicates its value. Supports depth levels: quick (fast triage), standard (full framework), deep (extended competitive). Produces structured context files (.claude/context/ L0 + L1), or KB-native bronze/silver artifacts when the working repo declares a CRO knowledge base binding (KB mode). Runs autonomous research by default. Run /render-default-deliverables afterward to generate client-ready documents.
tools1
aa-audit
When the user wants to audit Adobe Analytics data for a property. Also use when the user mentions 'AA audit,' 'Adobe Analytics audit,' 'AA performance profile,' or 'AA traffic analysis.' Runs a Python script against the AA 2.0 Reporting API, interprets the JSON output, and produces a structured performance-profile.md context file (.claude/context/ L1). Single agent, no depth flag. Works with any AA implementation given a client config file.
tools1
experiment-mockup
When the user wants to create a visual mockup of a proposed experiment change. Also use when the user mentions 'experiment mockup,' 'mockup hypothesis,' 'inject change,' 'DOM injection,' 'visual mockup,' 'mock up experiment,' 'show proposed change,' 'experiment preview,' or 'mockup for hypothesis N.' Takes a hypothesis from experiment-roadmap.md, navigates to the target page, injects the proposed change styled to match the site, iterates with the user, and captures the approved state as a standalone HTML artifact with CRO placement rationale. Three modes: live (Chrome DevTools MCP, interactive), playwright (Playwright MCP, screenshot-based iteration), and static (HTML extraction fallback, non-interactive).
tools1
landing-page-generator
When the user wants to generate a B2B paid landing page from existing positioning context. Also use when the user mentions 'landing page,' 'LP generator,' 'campaign page,' 'paid landing page,' 'landing page copy,' 'hero section,' or 'conversion page.' Four-phase pipeline with signal-driven section assembly: brief builder, copy agent (composable section selection), design agent, QA validator. Consumes L0+L1 context files from .claude/context/ and produces campaign deliverables in .claude/deliverables/campaigns/.
development1
positioning-update
When the user wants to apply client feedback, stakeholder corrections, or new intelligence to existing positioning context files. Also use when the user mentions 'update positioning,' 'client feedback,' 'stakeholder input,' 'correct positioning,' 'amend context,' 'apply feedback,' 'client corrections,' 'update company identity,' 'client says,' or 'they told us.' Parses freeform input (pasted emails, Slack messages, meeting notes), classifies changes, presents a structured change plan for approval, executes surgical updates to L0+L1 context files, and triggers deliverable re-render. No web research. Amendment skill, not research skill.
tools1
render-default-deliverables
When the user wants to generate client-ready deliverables from existing positioning context. Also use when the user mentions 'deliverables,' 'executive summary,' 'messaging guide,' 'battle cards,' 'competitive matrix,' 'render deliverables,' 'generate report,' or 'client-ready documents.' Reads L0 + L1 context files from .claude/context/ and produces polished, human-readable documents in .claude/deliverables/. No research, no analysis, no web fetches. Pure synthesis and formatting.
tools1
voice-inference
When the user wants to analyze a company's brand voice from its website content. Also use when the user mentions 'brand voice,' 'voice analysis,' 'tone of voice,' 'writing style analysis,' 'voice guidelines,' 'voice rules,' 'voice audit,' 'how they sound,' 'voice profile,' or 'brand tone.' Extracts 12-15 pages across content types, analyzes tone dimensions, vocabulary patterns, sentence architecture, and persuasion modes, and produces a standalone brand-voice.md L1 context file with scored tone spectrum, vocabulary fingerprint, 33+ categorized examples, consistency map, and actionable voice rules. Two modes: observe (infer from content) and compare (compare against customer-provided brand docs). Auto-detects brand docs in context directory. Does NOT require positioning-framework to have been run first.
development1
ga4-audit
When the user wants to audit GA4 analytics data for a property. Also use when the user mentions 'GA4 audit,' 'analytics audit,' 'traffic analysis,' 'page performance,' 'conversion audit,' 'bounce rate analysis,' or 'performance profile.' Pulls 11-15 targeted reports from GA4 via direct API or analytics-mcp fallback (including element-level interaction discovery and AI-referrer traffic segmentation), classifies events, and produces a structured performance-profile.md context file (.claude/context/ L1). Single agent, no depth flag. Works with any GA4 property.
tools1