CRO, Research & Reporting Toolkit

A strategist's toolkit for Claude (works in Claude, Cowork, and Claude Code). It packages reusable competencies distilled from real conversion-optimization engagements and from a study of the best public skills, agents, and processes on GitHub: audit a conversion funnel, run deep sourced research, write conversion copy, position an offer, and turn findings into polished, render-safe reports and redesign blueprints.

What's new in 0.3.0

• A real "begin here" entry point: a getting-started skill + a /kickoff command that orient you, capture a reusable engagement brief (./cro-engagement.md), and route to the right skill/command/agent.
• Shared context across the suite: every skill now reads the engagement brief first, so the tools stay consistent and stop re-asking (the coreyhaines31 shared-context pattern).

What's new in 0.2.0

• Two new skills: conversion-copywriting and outbound-and-positioning.
• Four specialized subagents: cro-auditor, deep-researcher, report-builder, cro-verifier (an adversarial pre-ship checker).
• Marketplace install: ships .claude-plugin/marketplace.json so it can be added via /plugin marketplace add.
• Deeper CRO references: experimentation rigor (A/B sizing, peeking, sequential testing, ICE/PIE/PXL), checkout/mobile/accessibility (Baymard/GoodUI/NN-g), persuasion & mental models (Cialdini + pricing), and a Shopify Liquid build checklist (current Dawn/theme-tools practice).
• Sharper research & reporting: a research-rigor reference (reflection/gap loops, credibility scoring, stop conditions) and a citations-and-grounding reference (gapless citation spec, outline-first generation, grounding gate).
• Tighter skill descriptions (triggers without leaking the method), license + metadata on every skill, and argument-hint + scoped allowed-tools on every command.

Components

Skills (auto-trigger on matching requests)

• getting-started — the front door: orients you, captures a reusable engagement brief, and routes to the right skill/command/agent. Begin here (or run /kickoff).
• ecommerce-cro-audit — diagnose why a page/funnel underconverts; benchmarks by tier, a five-gate diagnostic, experimentation rigor, premium psychology, Shopify hooks, impact/effort prioritization. Extends to landing/lead-gen/SaaS.
• deep-research-synthesis — thorough, multi-source, data-grounded research with source-reliability tiering, adversarial verification, and reflection/stop discipline.
• report-information-design — structure findings (BLUF), build self-contained HTML with render-safe components, cite with a gapless spec, and QA before delivering.
• conversion-copywriting — PDP/landing/email/ad copy with proven formulas, the 4 U's headline test, and a clear non-hype voice.
• outbound-and-positioning — cold email + cadence, B2B qualification (SPIN/Challenger/MEDDIC/JTBD), and positioning (April Dunford + Value Proposition Canvas).

Agents (delegated, auto-invoked where relevant)

• cro-auditor (read-only, Opus) — runs the audit and returns impact/effort-ranked findings.
• deep-researcher (Opus) — a self-contained worker for one parallel research stream.
• report-builder (Sonnet, writer) — assembles and QAs the report.
• cro-verifier (Opus, read-only) — verifies load-bearing claims, citation integrity, and that recommendations follow from evidence; reports "no findings" honestly.

Commands (explicit / entry points)

• /kickoff [what you want to do] — begin here: set up the engagement brief and route.
• /cro-audit <url|store|theme path> [--tier] [--funnel] [--audience]
• /deep-research <topic> [--depth] [--streams]
• /build-report <topic|files> [--format] [--audience]

How they work together

Begin with `/kickoff` to set up the engagement brief (./cro-engagement.md) that every skill then reads first. A typical engagement: /cro-audit grounds in first-party data, fans out deep-researcher for fresh evidence and cro-auditor for the page diagnostic, then report-builder assembles the deliverable and cro-verifier checks it before it ships. conversion-copywriting and outbound-and-positioning supply the on-page and outreach language. Each skill's "Related skills" section lists its neighbors.

Setup

No configuration required. More powerful when connected (all optional):

• Store analytics (e.g., a Shopify MCP with ShopifyQL) — grounds audits in real funnel data.
• Web search / fetch — powers the research skills.
• A codebase/theme export (file access) — maps findings to real files and supports the Liquid build checklist.

Install

• Marketplace: /plugin marketplace add Chykalophia/cro-research-reporting then install cro-research-reporting.
• Direct: save the .plugin file via the install button, or place the folder in your plugins directory.

Testing & iteration

Iterate skills with the official skill-creator harness pattern: for each skill, keep a small set of should-trigger and should-not-trigger prompts (the near-miss negatives matter most, since the skills have overlapping surfaces), test across Haiku/Sonnet/Opus, and refine the description until triggering is reliable. Keep each SKILL.md under ~500 lines and references one level deep.

Design notes & credits

• Evidence over assertion: every recommendation ties to data or a tiered source; vendor lift figures are labeled directional.
• Render-safe by default: report components encode hard-won fixes (no text inside data-width bars; no absolute-positioned label clouds; everything collapses on mobile; QA before shipping).
• Brand-faithful: reports match the subject's real brand tokens.
• Informed by public work (frameworks are public methods; MIT-licensed repos adapted with attribution): Anthropic Agent Skills (anthropics/skills), wshobson/agents and coreyhaines31/marketingskills (agent/skill patterns), gpt-researcher / Stanford STORM / LangChain open_deep_research (research rigor), Baymard Institute / Nielsen Norman Group / GoodUI / CXL-Speero (CRO knowledge), and Shopify Dawn / theme-tools (Shopify practice).

Version 0.3.0 · by Chykalophia.