One Claude skill for your Shopify and Klaviyo ops from chat
Claude skill that reads Shopify and Klaviyo in the same pass
A churn risk analysis usually needs two separate lookups.
Klaviyo for who's actually at risk and how much it's worth. Shopify for what they buy, whether it's even in stock, and what to lead with in the email.
Then stitch the two answers into a decision by hand.
I built a Claude skill that skips the stitching.
What a skill actually is
The prompt packs I've shared before this were built to be copied.
A Claude Skill works differently. It's a folder of markdown, not code, that installs into Claude once. After that, you don't paste anything.
With this, you can type what you want in plain language - it matches your request to one of 62 prompts, fills in the variables from what you asked, and runs it. Type something ambiguous, it won't guess - asks you to choose from the closest matches.
62 prompts split across 9 domains
1/ Catalog & Taxonomy (12 prompts) - the biggest section.
Attribute gaps, duplicate listings, taxonomy mapping, category mismatches - the reasons a product goes unmatched before you even reach a sales question.
2/ PDP & Content (5 prompts) - human-vs-agent page teardowns, structured content blocks, description rewrites that cut the adjectives and keep the facts.
3/ Scoring & Segmentation (6) - the churn and value scores behind the example above, plus a segment audit that flags every dead relic still sitting in Klaviyo unused.
4/ Retention & Lifecycle (7) - winback waves, cadence-break triggers, a holdout test built in so a program has to prove its lift instead of assuming it.
5/ Email & Flows (10) - the second largest domain. Winback emails, replenishment nudges, subject line matrices, mechanism-first, no filler adjectives.
6/ Acquisition & Ads (7 prompts) - lookalike seeds built from real top customers, Meta and Google ad sets, TikTok hooks.
7/ Analytics & Reporting (3 prompts) - the weekly scorecard, product ranking, anomaly scan, same shape every time.
8/ Merchandising, Collections & Pricing (6 prompts) - collection gaps, pricing ladder reviews, margin-aware promo planning.
9/ Agent Readiness & Audits (6 prompts) - a full-brand readiness scorecard, a schema audit, a fix list ranked by impact over effort.
What stays true across all the prompts
Every prompt reads by default.
If a prompt would write back to Shopify or Klaviyo instead of just reading it, that takes a separate, explicit instruction and a shown plan before anything runs.
Each one also ships with a fixed output contract, the exact shape the answer comes back in. Run the same prompt twice, a week apart, or hand it to someone else on the team, and the answer comes back in the same shape both times.
Getting it running
Works in Claude Chat, Claude Code, and Claude Desktop on Free, Pro, and Team plans. Team and Enterprise admins can install it once for the whole workspace.
In Claude.ai → Settings → Customize → Skills → Add, upload the file.
In Claude Code → unzip it into ~/.claude/skills/ so SKILL.md sits at ~/.claude/skills/the-agentic-operator/SKILL.md, then restart.
Get the skill for free: theagenticoperator.co/skills/the-agentic-operator
Tell me if a prompt returns something different from what's in the pack, or if any of the domains can have more. I'll keep building on it.
- Ankit
If this was useful, the next one will be too.
Weekly. Free.
Unsubscribe with one click.
