How to Use Claude for Lead Generation: A Step-by-Step Guide: The Outbound Operator's Playbook

Last Updated: May 11, 2026

This guide covers how to use Claude for lead generation end-to-end: sourcing accounts, scoring them against your ICP, enriching contacts, writing copy, and deploying campaigns, all from a single Claude Code workflow in your terminal.

Key Takeaways (TL;DR)

• Claude Code runs your entire outbound process, from building company lists to launching campaigns, without leaving your terminal.
• The 7 steps are: sourcing companies, scoring accounts, adding buying signals, finding contacts, enriching emails, writing copy, and deploying campaigns.
• Setup happens once: a CLAUDE.md file, a copy of the frameworks doc, a scoring criteria doc, and your API keys. After that, campaigns take minutes, not hours.
• The tool stack runs in layers: data (Apollo, Clay), enrichment (FullEnrich, Prospeo), signals (PredictLeads, Trigify), validation (BounceBan), and sending (Instantly, Lemlist).
• Our team of experts built and refined this system, running live client campaigns daily. The cheat sheets and frameworks in this guide come directly from that production experience.

Table of Contents

• Key Takeaways (TL;DR)
• How to Use Claude for Lead Generation: at a Glance
• Why Use Claude for Lead Generation?
• The One-Time Setup: What You Need Before Running
• How to Use Claude for Lead Generation: The Full 7-Step Process
• How Do You Choose the Right Claude Code Approach for Your Team?
• The Full API Stack by Layer
• Useful Claude Code Commands
• Everything You Need to Know About Claude Lead Generation
• Claude Code Gets You Started. A GTM System Gets You Results.
• Work With Frontal’s GTM Team
• Getting Started in Three Steps
• FAQs About How to Use Claude for Lead Generation

How to Use Claude for Lead Generation: at a Glance

If you want to use Claude for lead generation, the workflow runs across 8 phases: building a company list, scoring accounts, layering signals, finding decision-makers, enriching contacts, writing copy, deploying campaigns, and optimizing on the results.

Phase

What Claude Does

Key Tools

Company List

Processes 50k+ rows from CSV, pulls from multiple sources

Apollo, Clay, Instantly SuperSearch, Openmart, LeadsFactory, BuiltWith

ICP Scoring

Runs scoring criteria against raw data, tiers into T1/T2/T3

Python scripts, CLAUDE.md scoring criteria doc

Buying Signals

Detects hiring changes, tech shifts, and social signals

RB2B, PredictLeads API, Trigify.io API

Decision-Makers

Builds search criteria from ICP, finds 4-5 contacts per account

LeadsFactory API

Enrichment

Multi-provider waterfall, cleans risky emails pre-send

Prospeo, CompanyEnrich, FullEnrich, BounceBan

Copywriting

Pulls winning templates, writes signal-based angles per persona

Instantly.ai API, Lemlist API, copy frameworks doc

Deployment

Creates campaigns, injects copy, uploads leads, and sets sending

Instantly.ai, Lemlist (via API)

Optimization

Analyzes metrics, finds top segments, auto-builds lookalikes

Campaign analytics from sequencer

Why Use Claude for Lead Generation?

It comes down to three problems that compound on each other:

• Manual outbound doesn’t scale. A good SDR can research, personalize, and send maybe 30-50 quality emails a day. Claude processes 50,000+ rows from a CSV and writes personalized, signal-driven sequences for each one in the time it takes your SDR to brew coffee.
• Generic outreach kills your domain. Spray-and-pray email has wrecked domain reputations across B2B. Claude runs a multi-provider enrichment waterfall, validates emails before they hit your sequencer, and sends in controlled batches, so you protect your sending infrastructure while getting better reply rates.
• Disconnected tools kill momentum. The average outbound stack has 8-12 tools that don’t talk to each other. Clay for enrichment. Apollo for data. Instantly for sending. PredictLeads somewhere on the side. Claude connects all of them into one workflow, and your folder structure becomes your GTM brain. The outbound teams building a full pipeline in 2026 aren’t hiring more SDRs. They’re building Claude-powered systems that get better with every campaign.

What Claude Code Automates

Here’s exactly what gets handled without you touching it:

Capability

What it does

ICP scoring at scale

Reads your scoring criteria doc, runs Python against raw data, outputs a tiered account list with zero manual review.

API chaining

Reads the API docs itself. Connects enrichment providers, sequencers, and CRMs without you writing a single line of code.

Copy generation

Pulls analytics from your sequencer, identifies best-performing templates, and injects context for the current campaign.

Campaign creation

Creates the campaign, adds the copy, uploads the leads, and sets all delivery settings directly inside your sequencer.

Enrichment waterfall

Routes each contact through multiple providers in sequence. Stops when a verified email is found. Validates before sending.

Context memory

Stores leads, copy, campaign data, and reply analytics in one place. Uses past performance to improve every future campaign.

[[a]](#cmnt1)

The One-Time Setup: What You Need Before Running

You only need to do this once. After it’s done, campaigns run on their own.

0. Install Claude Code

Claude Code runs in your terminal. No IDE required.

Platform

Command

Mac / Linux

curl -fsSL https://claude.ai/install.sh | bash

Windows

irm https://claude.ai/install.ps1 | iex

Homebrew

brew install —cask claude-code

1. CLAUDE.md: Your Operating System

CLAUDE.md is the file Claude Code reads before it does anything. Think of it as the briefing doc for your outbound system. Without it, Claude produces generic output. With it, it operates as an expert in your stack, your scoring rules, and your workflows.

CLAUDE.md has three scopes, each serving a different purpose:

File

Scope

Use for

./[CLAUDE.md]

Project (shared with team)

Campaign rules, ICP definition, tool preferences

~/.claude/[CLAUDE.md]

User (all projects)

Your personal defaults across every client

.claude/rules

Path-specific

Rules that apply only to a specific folder or workflow

Keep it under 200 lines. It should cover your ICP definition, lead scoring criteria, which enrichment provider to try first, copy framework references, and any workflow logic specific to how you run campaigns.

ICP Definition

Target: B2B SaaS companies, 20-500 employees, $1M+ ARR

Exclude: Agencies, non-English speaking markets, bootstrapped solopreneurs

Lead Scoring

Tier 1: CEO/Founder + hiring for Sales/Growth + Series A+ funding

Tier 2: VP Sales or Head of Marketing + 50+ employees

Tier 3: Everyone else who matches ICP

Tool Preferences

Email enrichment order: Prospeo > FullEnrich > CompanyEnrich

Validation: BounceBan (run before upload)

Sequencer: Instantly.ai

Copy Rules

Always pull from /skills/copy-frameworks.md

Match template to signal type (hiring = growth angle, tech change = displacement)

Batch max 200 leads per campaign for quality control

2. Skills Files: Reusable Knowledge Packs (Skills = Knowledge)

Skills live in your .claude/skills/ directory. Each one stores knowledge for a specific task: how to run enrichment waterfalls, which copy template to pick, and how to score leads against your ICP.

Skills tell Claude what to do. MCP Servers tell Claude how to do it. A skill sets the scoring criteria. An MCP connection to Apollo makes the actual API call.

To create a skill, add a SKILL.md file in your skills directory with frontmatter:

name: my-skill

description: When to use this skill

user-invocable: true

Claude auto-loads skills when relevant and only loads what’s needed per task, keeping sessions lean and on-task.

3. MCP Servers: Your App Store (MCP = Actions)

MCP Servers are how Claude connects to external tools. Once connected, they give Claude full access to act inside those platforms directly from the terminal.

• Remote: claude mcp add —transport http
• Local: claude mcp add —transport stdio Servers can be scoped to just you (local), shared with a team (project), or applied across all projects (user). Common integrations used in outbound workflows include Notion, GitHub, Slack, Gmail, and Airtable.

4. The Four Documents You Set Up Once

The cheat sheet breaks the one-time setup into four specific docs. Each serves a distinct role in the workflow:

Document

What it does

CLAUDE.md file

Master directions doc. Scoring rules, tool preferences, and workflow logic. Claude reads this before every task.

API keys

Connect your data, enrichment, validation, and sequencing tools. Paste the keys once, and Claude reads the docs and handles the rest.

Copy frameworks doc

Your best-performing templates. Claude pulls the right one per campaign based on goal and audience automatically.

Scoring criteria doc

ICP filters written in plain English. Claude runs Python against your CSV with zero guesswork, no hallucination.

5. API Keys

Paste in your API keys for every tool in your stack. Claude reads the docs and handles the connections. You do it once; Claude takes it from there.

How to Use Claude for Lead Generation: The Full 7-Step Process

Here’s what it looks like from a blank prompt to a running campaign:

Step 1: Build Your Company List

Claude Code can pull company data from several places at once. Two starting points that work well:

• Lookalike search via PredictLeads. Feed in your top 10 existing customers, and PredictLeads returns companies with matching signals. You’re cloning the accounts that already pay you.
• Direct pulls. Clay, Apollo.io, Sales Navigator, or Openmart all export into a CSV that Claude can pick up and run with. Claude processes 50,000+ rows without any manual cleaning. It reads the column headers, fixes inconsistent formatting, and has the data ready to score.

Clay is central to how this workflow runs. Our team of experts is one of only 4 Elite Studio Clay Partners globally, at the highest tier of Clay’s partner program. When this system is built for clients, they get Clay at full depth, not the surface-level setup most teams run for months before seeing results.

Tools: RB2B · PredictLeads · Clay · Apollo.io · Instantly SuperSearch · Sales Navigator · Openmart · LeadsFactory · BuiltWith

Step 2: Score and Tier Accounts

Claude runs your ICP criteria against the raw CSV data using your lead-scoring skill. No manual review. It reads the scoring rules in your CLAUDE.md, applies them row by row, and outputs a tiered list:

• Tier 1 — Your highest-value accounts. Worth a personalized 1:1 approach.
• Tier 2 — Multi-channel outreach (email + LinkedIn sequence).
• Tier 3 — Email only. Broad ICP match, less personalization investment. The tiering matters because it decides how much enrichment and copy budget each segment gets. You won’t spend Tier 1 effort on Tier 3 accounts.

Tools: Scoring criteria doc · CLAUDE.md · Python script via Claude · CSV input

Step 3: Layer Buying Signals

Knowing who to target is one thing. Knowing when to reach out is another. Claude calls these APIs to add timing data on top of your scored accounts:

• PredictLeads API. Picks up hiring activity, tech stack changes, product launches, and press coverage. If a company just hired 10 SDRs, they’re scaling sales. That’s your opening.
• Trigify.io API. Spots social signals: LinkedIn comments, competitor mentions, and content engagement. If a prospect is commenting on a competitor’s post, they’re already thinking about the problem you solve.
• RB2B. Identifies anonymous website visitors and matches them back to named accounts, so you can reach out to people who’ve already shown interest before you’ve said a word. These signals flow directly into Step 6, so the email Claude writes actually references what’s happening at that company right now.

Tools: RB2B · PredictLeads API · Trigify.io API

Step 4: Find Decision-Makers

Claude reads your ICP definition from CLAUDE.md, including job titles, seniority, department, and geography, and calls LeadsFactory’s API to find 4-5 decision-makers per company.

Instead of finding one contact and hoping they reply, you get multiple people at each account: the economic buyer, the end user, the internal champion. Each one gets a different angle in the campaign.

Tools: LeadsFactory API · CLAUDE.md ICP definition

Step 5: Enrich and Validate Contacts

Claude runs a waterfall across multiple enrichment providers to find a verified email for every contact. Try the first provider; if there’s no result or the address looks risky, move to the next. You get maximum coverage without paying for the same lookup twice.

• Email enrichment waterfall: Prospeo > CompanyEnrich > FullEnrich
• Validation: BounceBan removes risky addresses before anything reaches your sequencer This is what protects your domain. Sending to unvalidated lists pushes bounce rates up and deliverability down. Frontal’s campaigns consistently hit 0.5% bounce rates, well below the industry threshold, because validation is built in, not bolted on.

Tools: Prospeo · CompanyEnrich · FullEnrich · BounceBan

Step 6: Generate Copy and Deploy

Claude pulls your best-performing copy templates from Instantly.ai or Lemlist via API, then writes a specific angle for each lead based on:

• The signal from Step 3 (hiring push, tech change, social activity)
• What’s visible on the company’s public-facing presence right now
• Who you’re emailing (a VP of Sales needs a different opener than a Head of Marketing) Batches cap at 100-200 leads. Once they’re written, Claude creates the campaign inside your sequencer, drops in the copy, uploads the lead list, and sets the sending configuration, all via API. You don’t touch it.

Why batch size matters: Sending 2,000 leads in one shot degrades reply rates and kills your ability to iterate. At 100-200, you can test angles, see what’s working, and scale only what earns it.

Tools: Instantly.ai API · Lemlist API · Copy frameworks doc

Step 7: Gather Feedback and Auto-Improve

Most outbound tools stop once the campaign goes live. Claude Code doesn’t. After launch, it pulls your campaign metrics, breaks down performance by segment, and feeds what it learns back into the next run:

• Checks reply rates, positive reply rates, and booked meetings across segments and message variants
• Spots which Tier 1 accounts replied and builds a lookalike list for next month
• Updates your scoring rules when certain signals consistently lead to higher reply rates Think of it like a retargeting pixel for outbound. Each campaign teaches the system something. Results compound.

Tools: Instantly.ai analytics · Lemlist analytics · Scoring criteria doc · Context memory

Useful Claude Code Commands to Keep Your Campaigns Running Cleanly

Once your core workflow is set up, these commands save time and keep sessions from going sideways. None of them are required for your first campaign, but you’ll reach for them regularly once you’re running at scale.

Slash Commands

Command

What it does

/init

Generates your CLAUDE.md automatically from your existing project folder

/memory

Persistent memory across sessions

/review

Reviews PRs or current session work

/rename

Names the current session

/compact

Compresses context when a long session starts eating your context window

/cost

Shows token usage for the current session

/config

Opens settings

/vim

Vim mode

You can also build custom commands from your own Skills files, turning repeatable workflows into one-line prompts.

Session Management Commands

Command

What it does

/rename

Name the current session

claude -c

Continue last session

claude —resume

Browse all sessions

claude -n

Start with a name

claude -r “name”

Resume by name

—fork-session

Branch a session without affecting the original

Name every session. Resume in seconds.

Keyboard Shortcuts

Shortcut

Action

Ctrl+C

Cancel

Ctrl+R

History search

Esc Esc

Rewind

Alt+T

Extended thinking

Ctrl+L

Clear screen

Ctrl+G

Open in editor

Alt+P

Switch model

Ctrl+V

Paste images

Opt+Enter (Mac) / Backslash+Enter (Universal)

Multiline input

CLI Power Moves

claude -p “query” # One-shot mode

claude —model opus # Pick model

claude —max-turns 5 # Limit loops

claude —add-dir …/repo # Extra context

cat file | claude -p # Pipe input

claude —effort high # Max reasoning

claude —max-budget-usd 5 # Spending cap

claude —permission-mode plan # Read-only

Headless mode (-p) = scripts, CI/CD, automations.

Hooks: Automations That Run on Their Own

Hooks fire at specific points in Claude’s workflow without you having to ask:

Hook

What it does

PreToolUse

Block dangerous commands before they execute

PostToolUse

Auto-format code after Claude writes it

SessionStart

Re-inject context at the start of every session

Stop

Verify task completion before Claude exits

Notification

Desktop alerts when Claude needs your attention

Set in .claude/settings.json.

Model Switching

Use /model mid-session to switch between Claude models:

• Haiku for fast, repetitive tasks like data normalization or list formatting
• Sonnet or Opus for copy generation and scoring logic where output quality matters

[[b]](#cmnt2)

How Do You Choose the Right Claude Code Approach for Your Team?

You don’t need to run all 7 steps on day one. Where you start depends on what’s actually broken.

• If your data’s messy, start with Steps 1-2. Get Claude processing your existing lists, cleaning up bad CSVs, and scoring accounts against your ICP. That alone cuts hours of manual work and gives you a cleaner foundation before you spend anything on enrichment.
• If you’re sending but getting no replies, the problem is usually relevance, not volume. Go to Steps 3 and 6: signal detection and signal-based copy. When your emails reference what’s actually happening at each company, they stop looking like templates.
• If you’ve got too many tools that don’t talk to each other, start with CLAUDE.md and your API connections. You probably don’t need new tools. You need Claude to coordinate the ones you already have.
• If you’re starting from scratch, go through the 7 steps in order. Each one feeds the next. Jumping straight to deployment without enriching first is how you burn a domain.
• If you’re running an agency, the skills system is where the real value is. Each client gets their own CLAUDE.md and skills files. The infrastructure stays the same; the rules and inputs change. Claude runs the execution; your team owns the strategy and the client relationship. Most teams should complete the one-time setup, including CLAUDE.md, two or three skills files, and existing API keys, then run one campaign from list to deployment before automating anything. You’ll understand the system much better once you’ve seen each step work.

The Full API Stack by Layer

Here’s every tool broken out by what it does in the workflow. You don’t need all of them on day one. Pick one provider per layer and expand once you’ve seen what works.

Layer

Function

Tools

Signal

Where intent lives

RB2B, PredictLeads, Trigify

Data — Find

Where companies and contacts get built

Apollo.io, Instantly SuperSearch, Openmart, LeadsFactory, BuiltWith

Data — Enrich

Find and verify email addresses

FullEnrich, CompanyEnrich, Prospeo

Data — Validate

Remove risky addresses before sending

BounceBan

Automation

Coordinate workflow steps

Clay, n8n, Relevance AI

Action

Where campaigns go live

Instantly.ai, Lemlist

Everything You Need to Know About Claude Lead Generation

Topic

Summary

What it is

Using Claude Code to automate the full outbound lead generation process: sourcing, scoring, enriching, writing, and deploying, via API integrations and a structured skills system.

Core requirement

CLAUDE.md (operating file), skills docs (reusable knowledge), and API keys connected to your data, enrichment, and sequencing stack.

Data scale

Claude Code handles 50,000+ rows from CSV files. It normalizes messy data and prepares it for scoring without manual cleaning.

Scoring

Runs Python-based ICP scoring against raw data. Outputs Tier 1 (manual), Tier 2 (multi-channel), and Tier 3 (email-only) segments.

Signal sources

PredictLeads for hiring, tech, and product signals. Trigify.io for social signals. Both feed directly into copy generation.

Enrichment method

Waterfall approach: Prospeo > CompanyEnrich > FullEnrich. BounceBan validates all addresses before deployment.

Copy generation

Pulls best-performing templates from sequencer via API, writes signal-based angles per persona, batches at 100-200 leads for quality control.

Campaign deployment

Fully autonomous via API. Creates campaign, injects copy, uploads leads, and configures sending settings inside Instantly.ai or Lemlist.

Optimization loop

Analyzes post-launch metrics, identifies top-performing segments, builds lookalikes from best leads, and updates scoring rules. Campaigns improve automatically.

Time to first campaign

Initial setup takes a few hours. First campaign after setup: minutes. Frontal clients go live within 2 weeks of signing, not 6-8 weeks like most agencies.

Who this suits

B2B SaaS founders, GTM operators, RevOps leads, and growth engineers who want to scale their pipeline without scaling headcount linearly.

Claude Code Gets You Started. A GTM System Gets You Results.

Knowing how to run Claude Code and knowing how to build a consistent pipeline are two different things. This guide gives you everything you need to build a Claude Code outbound workflow. But there’s a gap worth naming before you close the tab.

• The tool executes what you tell it. Claude Code is only as good as the thinking behind it. If your ICP is too broad, your scoring criteria are off, or your copy angles don’t match the signals you’re pulling, the automation scales the mistake. Getting the strategy right first isn’t optional. It’s the whole game.
• Outbound in isolation doesn’t compound. Claude Code handles cold email and LinkedIn outreach well. But the accounts that convert to pipeline have usually seen your content, been touched by ads, or recognized your brand before your email landed.
• The feedback loop in Step 7 still needs judgment. It pulls the data, but someone has to read it. Which reply rate improvements actually matter? Which segments are worth scaling? When does a shift in signals mean you need to rethink your ICP, not just your subject line? Those are calls that can’t be delegated to a prompt. That’s the gap between having the tools and actually running the system. Our team of experts has built and run this system daily across live client accounts, not as a test environment, but as production infrastructure. We know what breaks, how to fix it fast, and how to make the whole thing compound instead of plateau.

If you’re a B2B SaaS founder or VP of Growth doing $100K+ per month and you want a pipeline, not just a configured workflow, here’s what working with Frontal looks like.

Work With Frontal’s GTM Team

Running a Claude Code outbound system well takes time, real knowledge of the stack, and enough iteration to know what breaks and why. Our team of experts runs this system daily across live client accounts. This isn’t a demo environment. It’s how we build a pipeline for B2B SaaS teams every day.

Here’s what the system produces:

• AirOps — $7.83M in qualified pipeline, $1.52M closed-won revenue, 164 deals created in 10 months

• Scaling: Tens of thousands of prospecting meetings booked across 500+ GTM teams

• Organic LinkedIn: 30M+ organic views generated for clients and brand What sets us apart:

• 3-channel GTM Flywheel — Claude Code outbound runs alongside LinkedIn Ads (ABM) and LinkedIn Content as one coordinated system. Each channel compounds on the others. Most agencies run one channel. We run three.

• Elite Studio Clay Partner — one of only 4 globally at the highest tier of Clay’s partner program. Advanced personalization at a depth most agencies can’t replicate.

• Signal-driven, account-based execution — every action is triggered by intent data. No spray-and-pray.

• First campaigns live in 2 weeks — not 6-8 weeks like most agencies.

• Full-transparency reporting — live weekly dashboards covering deliverability, engagement, pipeline, and infrastructure health. No black box.

• Risk-reversed pilot — the first 90 days are a structured pilot. If results aren’t there at Day 90, you keep the playbook, assets, and infrastructure. No lock-in. Book a free strategy call with Frontal today and let us build your revenue engine.

FAQs About How to Use Claude for Lead Generation

Can you use Claude AI for lead generation?

Yes. Claude AI can run the full lead generation process when you pair it with Claude Code and the right API stack. It processes 50,000+ rows of CSV data, scores accounts against your ICP, finds decision-makers, enriches contacts, writes personalized copy, and deploys campaigns via Instantly.ai or Lemlist, all without leaving the terminal. The key is getting CLAUDE.md and your skills files set up before you run anything.

What is CLAUDE.md, and why does it matter for outbound?

CLAUDE.md is the file Claude Code reads before it does anything. For outbound, it holds your ICP scoring rules, tool preferences, workflow logic, and copy frameworks. Without it, Claude produces generic output. With it, it applies your exact methodology every single run. Keep it under 200 lines and update it as your stack changes.

How does Claude Code compare to traditional SDR tools for lead generation?

Claude Code takes on the work that used to require a person moving data between platforms: sourcing, enrichment, copywriting, and campaign setup. It does this by chaining APIs together, ingesting CSV data, running ICP scoring, calling enrichment providers, and pushing campaigns directly into your sequencer. Traditional SDR tools need a person at every handoff. Claude Code connects those handoffs automatically.

What APIs does Claude Code use for lead generation?

The stack works in layers. For data: Apollo.io, Clay, Openmart, and LeadsFactory. For enrichment: FullEnrich, CompanyEnrich, and Prospeo. For signals: PredictLeads (hiring, tech stack changes) and Trigify.io (social signals). For validation: BounceBan. For sending: Instantly.ai and Lemlist. Automation runs through Clay, n8n, or Relevance AI, depending on how your setup is configured.

How long does it take to set up a lead generation workflow in Claude Code?

The first-time setup, including CLAUDE.md, skills files, and API keys, takes a few hours if you’ve got RevOps or growth engineering experience. After that, each campaign takes minutes to build. Frontal clients typically get their first campaigns live within 2 weeks of starting, which covers both infrastructure setup and campaign prep.

What is Claude lead generation, and how does it work?

Claude lead generation means using Claude Code to run your outbound prospecting from start to finish. It reads your ICP and workflow instructions from a CLAUDE.md file, then works through 7 steps: building a company list, scoring accounts, adding buying signals, finding decision-makers, enriching contacts, writing copy, and deploying campaigns via your sequencer’s API. Each campaign also feeds data back into the next one, so results improve over time.

Does Claude AI lead generation work without coding knowledge?

You’ll need to be comfortable with terminal-based tools and API keys, but you don’t need to write code from scratch. The setup involves writing CLAUDE.md, creating skills files, and pasting in API credentials. Teams with RevOps or growth engineering backgrounds pick it up fastest. The learning curve is front-loaded; once it’s configured, you’re working with prompts. Our team of experts has published cheat sheets and frameworks, referenced throughout this guide, that cut the setup time considerably.

Can Claude replace a human SDR for outbound lead generation?

Claude Code handles the research, enrichment, copywriting, and campaign setup that SDRs typically spend 80% of their time on. What it can’t do is run a discovery call, navigate a procurement process, or read a relationship. Claude Code handles the infrastructure, so your reps can focus on the conversations that actually close deals.

What are Skills in Claude Code, and how do they help with outbound?

Skills are reusable knowledge docs stored in your .claude/skills/ directory. Each one tells Claude how to handle a specific task: enrichment waterfalls, ICP scoring logic, which copy template to use for which campaign type. When Claude Code starts a task, it pulls in the relevant skill automatically. Skills carry the knowledge; MCP Servers make the API calls. For outbound, you’d typically maintain separate skills for enrichment, copy, and scoring flows.