What are Claude Skills?

A skill is a set of instructions, packaged as a simple folder, that teaches Claude how to handle specific tasks or workflows.

A skill folder contains:

• SKILL.md (required): Instructions in Markdown with YAML frontmatter

• scripts/ (optional): Executable code (Python, Bash, etc.)

• references/ (optional): Documentation loaded as needed

• assets/ (optional): Templates or files used in output

Skills use a three-level system:

• First level (YAML frontmatter): Always loaded. Tells Claude when to use the skill.

• Second level (SKILL.md body): Loaded when Claude thinks the skill is relevant. Contains full instructions.

• Third level (Linked files): Additional files Claude can navigate and load as needed.

This keeps token usage low while giving Claude deep, specialized knowledge exactly when it needs it.

What is Lemonado and where does it fit?

Lemonado is an AI-native data platform for marketers. It unifies your ad platforms, analytics tools, and business data into a single queryable layer so AI agents can answer questions, automate workflows, and monitor performance on your behalf.

Lemonado connects to Claude via MCP (Model Context Protocol), the same open standard that Skills are built on.

The kitchen analogy from Anthropic says it well:

• MCP provides the professional kitchen: access to tools, ingredients, and equipment

• Skills provide the recipes: step-by-step instructions on how to create something valuable

Lemonado is the kitchen, Skills are the recipes. Together they give Claude everything it needs to analyze, report on, and surface insights from your ad accounts.

Here's our full guide on how to connect your Lemonado workspace to Claude.

MCP without Skills vs. MCP with Skills

Who this guide is for:

• Performance marketers already using Claude who want to stop pasting CSVs

• Paid media managers running the same reports and audits every week

• Agency operators managing multiple clients who need consistent, scalable workflows

What you'll need:

• A Lemonado account with at least one data source connected. Sign up here for free

• A paid Claude subscription (Skills require Claude Pro or Team)

Three use cases to start with

Use this framework to identify where Skills will save you the most time.

Ask yourself:

• What do I explain to Claude over and over?

• What reports do I rebuild manually every week?

• What analysis do I run across every client account?

Those are your Skill candidates. Here are three to start with:

Use Case 1: Weekly Performance Report

Every week, the same task. Pull the numbers, calculate the deltas, write the summary, send it to the client. A Skill automates the workflow. Lemonado provides the live data it runs on. The result is a client-ready report that generates itself.

Trigger: "Run the weekly report for [client]" or "generate this week's performance summary"

Steps:

1. Query the last 7 days of campaign data from Lemonado for the specified client

2. Calculate core metrics: spend, ROAS, CPA, CTR, and conversions vs. the prior 7 days

3. Flag any campaign where performance moved more than 15% in either direction

4. Structure output as a client-ready summary: headline metrics, notable changes, and one recommended action per flagged campaign

Result: A formatted performance summary pulled from live data, ready to send without editing.

Key techniques:

• Dynamic date ranges so the Skill never uses hardcoded periods

• Client scoping via Lemonado so the query only pulls the right account

• Threshold-based flagging to surface what matters without noise

• Output templated to match your reporting format

Use Case 2: Account Audit

Account audits are high-value and time-consuming. They require pulling data across campaigns, normalizing it against benchmarks, and forming a clear point of view on what to fix first. This Skill runs that process end to end, coordinating multiple Lemonado queries in sequence before producing a prioritized output.

Trigger: "Audit [client]'s account" or "what needs attention this week" or "where are we losing money"

Steps:

1. Pull campaign and ad set data across all active spend from Lemonado

2. Score each campaign against configurable benchmarks: ROAS floor, CPA ceiling, CTR threshold, impression share minimum

3. Rank campaigns from worst to best performing against those benchmarks

4. For the bottom three, surface the specific metric that's failing and why it matters

5. Recommend a next action for each: pause, restructure, adjust bid strategy, or test new creative

Result: A prioritized audit with clear reasoning and recommended actions, grounded in live account data.

Key techniques:

• Multi-step MCP queries coordinated in sequence (campaigns, then ad sets, then creatives)

• Benchmark thresholds embedded in the Skill so you don't re-explain them every time

• Reasoning tied to specific metrics, not generic observations

• Configurable per client based on their goals and KPIs stored in Lemonado

Use Case 3: Creative Fatigue Review

Creative fatigue is one of the most common and most missed performance drains in paid media. CTR drops gradually, CPC creeps up, and by the time you notice, you've wasted two weeks of budget. This Skill monitors trend data across your active creatives and flags what needs to be replaced before it becomes a problem.

Trigger: "Check creative performance" or "what ads are burning out" or "which creatives should I refresh"

Steps:

1. Pull CTR and CPC data by creative for the last 14 days from Lemonado, segmented by ad set

2. Calculate week-over-week trend for each creative: CTR delta and CPC delta

3. Flag creatives where CTR has dropped more than 10% or CPC has risen more than 15% week-over-week

4. Identify the specific ad sets where fatigue is most concentrated

5. Cross-reference against top performers in the same campaign to surface what's working as a reference point for replacement

Result: A creative fatigue report that tells you exactly what to kill, where to focus new production, and what benchmarks to beat.

Key techniques:

• Trend analysis over rolling windows rather than single-period snapshots

• Ad set level segmentation so fatigue is caught before it drags campaign-level numbers

• Cross-referencing winners and losers in the same query for immediate creative direction

• Embeds paid media best practices around frequency thresholds and fatigue signals

Three community Skills to try now

These are built by Corey Haines and available as copy-paste markdown files. Each one becomes significantly more powerful when Lemonado MCP is connected. Instead of working from pasted data, Claude queries your live accounts for you.

1. paid-ads → View on GitHub 🔗

Covers campaign structure, optimization logic, and performance analysis for Google and Meta. With Lemonado connected, every recommendation is backed by live spend, ROAS, CPA, and impression share.

Install:

npx add-skill coreyhaines31/marketingskills --skill paid-ads

npx add-skill coreyhaines31/marketingskills --skill paid-ads

npx add-skill coreyhaines31/marketingskills --skill paid-ads

2. analytics-tracking → View on GitHub 🔗

Covers measurement, attribution, and conversion analysis. With Lemonado's Google Analytics integration connected alongside your ad platforms, this skill can answer the question most marketers can't easily answer: did the campaign drive conversions, or just clicks?

Install:

npx add-skill coreyhaines31/marketingskills --skill analytics-tracking

npx add-skill coreyhaines31/marketingskills --skill analytics-tracking

npx add-skill coreyhaines31/marketingskills --skill analytics-tracking

3. ad-creative → View on GitHub 🔗

Generates and iterates ad headlines, descriptions, and primary text. With Lemonado data behind it, Claude can see actual CTR and CPC performance per creative before writing the next iteration. It's not generating blindly — it's optimizing against what's actually working.

Install:

npx add-skill coreyhaines31/marketingskills --skill ad-creative

npx add-skill coreyhaines31/marketingskills --skill ad-creative

npx add-skill coreyhaines31/marketingskills --skill ad-creative

Build your own Lemonado-native Skill

Most community skills are general-purpose. A Lemonado-native Skill is scoped to your clients, your metrics, and your reporting structure.

Good starting points:

• A Budget Pacing Skill that checks mid-month spend across all active campaigns and flags clients projected to over or underspend

• A Brand vs. Non-Brand Skill that pulls search term data and splits performance automatically, no pivot tables required

• A Client Report Skill scoped to a specific account's goals, KPIs, and reporting tone

To build one, follow Anthropic's skill structure: create a folder, add a SKILL.md file with YAML frontmatter, and write your workflow as clear, step-by-step instructions. Reference your Lemonado MCP connection as the data tool.

Full build guide from Anthropic: resources.anthropic.com

Pre-launch checklist

Before your Skill is ready to use:

Setup

◽ Lemonado MCP connected and confirmed in Claude settings
◽ At least one data source active (Meta Ads, Google Ads, or Google Analytics)
◽ Paid Claude subscription active

If installing community skills

◽ Skill installed via npx add-skill or manual copy
◽ Tested with an obvious trigger phrase ("run a performance audit")
◽ Confirmed Claude queries Lemonado data, not asking you to paste anything

If building your own

◽ SKILL.md file exists with YAML frontmatter
◽ Trigger description covers WHAT and WHEN
◽ Steps are clear and reference Lemonado MCP as the data source
◽ Tested with real client data
◽ Error handling included (what if a data source returns nothing?)

Not ready to build? Start simpler.

If this feels like too much setup, you have two options.

Just want to ask questions first? Connect Lemonado MCP to Claude and ask: "How did my Meta campaigns perform last week?" No Skills required. The MCP connection alone gets you most of the way there.

Rather skip the setup entirely? Lemonado's built-in agent works out-of-the-box — no Skills to install, no MCP to configure, no Claude subscription required. You get:

• Live connectors for Meta Ads, Google Ads, LinkedIn Ads, Google Analytics, and more

• An AI agent that answers questions, builds reports, and audits accounts

• Scheduled reports delivered automatically: daily, weekly, or monthly

• Client-scoped dashboards so every client only sees their own data

• Multi-client management from a single workspace

We offer a 10-day free trial — signup here to get started.