Getting Started (begin here)

The front door. Orient the user, capture a reusable engagement brief, then route to the right tool. Run this whenever someone is starting, is unsure which skill applies, or asks something broad.

What this toolkit does (say this plainly)

A strategist's kit to: audit a conversion funnel, research customers, plan and read out tests, run deep sourced research, write conversion copy, sharpen positioning/outbound, and turn it into polished reports + redesign blueprints, with a verifier that checks the work before it ships. It uses AI where the evidence says AI is reliable (coding feedback, checking pages against fixed rules, drafting variants, running the maths) and keeps judgement calls with you.

Step 1 — Engagement brief (persistent context)

Before doing real work, establish the shared context every other skill reads first.

1. Look for an existing brief at ./cro-engagement.md (the default; or ask the user for the path). If present, read it, summarize it back in 2–3 lines, and ask if anything changed.
2. If absent and the task is more than a one-off, offer to capture one. Ask only what's needed (skip anything already obvious from the request), using references/engagement-brief-template.md:
   • Client / brand and what they sell; price tier (mass / premium / luxury).
   • ICP (who buys, what they value).
   • Surface in scope (PDP / collection / cart / checkout / landing / lead-gen / SaaS signup / whole funnel).
   • Primary goal & metric (e.g., add-to-cart rate, RPV, lead conversion).
   • Brand voice & tokens (tone, banned words, colors/fonts/logo if known).
   • Connected data/tools (analytics/Shopify MCP, web search, theme/codebase access), the testing tool, and traffic per key template.
   • Research assets (review export, tickets, past surveys, recordings) and the test log path.
   • Known constraints (timeline, what's off-limits, who the audience for deliverables is).
3. Write/update ./cro-engagement.md from the template. Keep it short and current; confirm before overwriting.

Why: the skills are most powerful when they share one context (so they stop re-asking and stay consistent). This is the coherence layer for the whole suite.

Step 2 — Route to the right tool

Match the user's intent to a component using references/routing-map.md. Quick version:

• "Why isn't this converting / audit this page or funnel" → /cro-audit or the ecommerce-cro-audit skill (it can delegate to the cro-auditor agent).
• "Research X thoroughly / with sources" → /deep-research or deep-research-synthesis (fans out deep-researcher agents).
• "Turn this into a report / write it up" → /build-report or report-information-design (or the report-builder agent).
• "What are customers saying / analyse reviews / survey questions / user test" → voc-research (bulk coding via the voc-coder agent).
• "Plan / size / read out a test", "how long should it run", "did it win" → /experiment or experiment-design.
• "Write/rewrite copy, headline, CTA, email, ad" → conversion-copywriting.
• "Cold outreach, follow-up cadence, positioning, value prop, qualify a deal" → outbound-and-positioning.
• "Check this before it ships" → the cro-verifier agent.

Step 3 — Show the recommended end-to-end flow

For a full engagement, the canonical path is: Brief (this skill) → /cro-audit (funnel data + rendered checklist + customer evidence → evidence-tagged hypotheses, test plans, report, verification) → copy variants → /experiment readout + test log. Offer to kick off the first step.

Reference map

• references/engagement-brief-template.md — the brief template, field definitions, where it lives, and update cadence.
• references/routing-map.md — full intent → skill/command/agent map and the end-to-end playbook.

Guardrails

Don't interrogate the user — ask only the brief questions that matter for the task at hand, and infer the rest. Don't overwrite an existing brief without confirming. If the user clearly already knows what they want, skip the orientation and route straight there.