Updated September 28, 2026 | 14 min read
Updated September 28, 2026 | 14 min read
Ready to launch?
Convert strategy into pipeline
Launch your first outbound campaign with verified leads, AI sequences, and multichannel outreach from one platform.
The Free Claude Skill Behind lemlist's 244K-Campaign Cold Email Framework (Copy the .md File)
<article>
I once watched a rep send 400 first-touch emails in a week and book nothing. The copy wasn’t lazy. He’d written every email himself, by hand, at night. The problem was that all 400 opened the same way, made the same ask, and read like the other eleven cold emails sitting in the same inbox that morning. When we built lemlist’s Outbound Analyst skill, it benchmarked campaigns against data from 244K+ real campaigns and 249M+ emails, and the same structural failures kept surfacing at the top of every underperforming sequence. Email one was almost always where it broke.
I once watched a rep send 400 first-touch emails in a week and book nothing. The copy wasn’t lazy. He’d written every email himself, by hand, at night. The problem was that all 400 opened the same way, made the same ask, and read like the other eleven cold emails sitting in the same inbox that morning. When we built lemlist’s Outbound Analyst skill, it benchmarked campaigns against data from 244K+ real campaigns and 249M+ emails, and the same structural failures kept surfacing at the top of every underperforming sequence. Email one was almost always where it broke.
So we wrote the rules down and turned them into a file Claude can follow. In this post, I’ll show you exactly how our Copywriting First Touch skill works: the doctrine it enforces, the subject line and opening line patterns it uses, how to install the .md file in under five minutes, and how to push the output into a live campaign. I’ll also cover the frameworks you could use instead, because a skill file isn’t the only way to write a decent first email.
Every way to write a first touch (and where a Claude skill fits)
Before you download anything, know what you’re choosing between.
You can write from a classic copywriting framework. AIDA (attention, interest, desire, action) and PAS (problem, agitate, solve) both predate cold email by decades and still hold up. Before-After-Bridge works well when your product changes a daily workflow. These give you a shape, but they say nothing about personalization, word count, or what to ask for on email one.
You can work from a swipe file. Save the emails that got you replies, keep a doc of subject lines that landed, and rebuild from there. This is the cheapest option and, honestly, a good one. It degrades the moment three reps start copying the same template into the same market.
You can write a long prompt in ChatGPT or Claude and paste it every time. It works until you tweak a word, forget which version you were using, and end up with two reps producing structurally different emails from what they think is the same instruction.
You can pull a skill from a public directory. Plenty exist now, including GTM-focused libraries like ColdIQ’s skills directory, and several of them are good.
A skill file solves the consistency problem specifically. It’s a Markdown document that sits in the conversation and constrains every draft Claude produces, so the framework doesn’t drift between prompt one and prompt forty. Anthropic published Agent Skills as an open standard built on files and folders, which is why a plain .md file works across clients instead of locking you into one tool. That’s the part I care about. The doctrine below is what makes our version different from a generic one.
What the copywriting first touch skill does
The Copywriting First Touch skill writes the opening email of a cold sequence: subject line, body, and CTA, calibrated to the prospect you describe. Think of it as handing Claude a playbook instead of asking it to improvise.
It activates when you ask Claude to write a first email, draft a cold email, or produce a single outreach message with no prior context. Each prompt returns one complete draft, plus a bridge line to email two if you ask for one. The file is free and open source, maintained in lemlist’s GitHub repo, and you can grab it from the lemlist Claude Skills Hub and have it running in a few minutes.
“First touch” means exactly what it says here. No prior relationship, no warm intro, no earlier thread to reference. That constraint changes the copy more than most reps expect, which is why the skill treats it as a separate job from follow-ups.
Why the first email in a sequence decides your reply rate
You get about three seconds. Sender name, subject line, the first line of preview text. That’s the whole audition. If those three seconds read like every other B2B pitch, the archive button wins and emails two through four never get a fair hearing.
Here’s the pattern most reps fall into without noticing: “I hope this finds you well,” the company name as the only personalization, then “do you have 15 minutes this week?” Prospects have pattern-matched that structure hundreds of times.
The failures below are the ones the skill’s rules exist to block. They’re the structural problems that keep showing up when you benchmark weak sequences against lemlist’s campaign data.
What kills the first touch | Why it fails |
|---|---|
“I hope this finds you well” opener | Burns the only line the prospect reads in preview, says nothing about them |
Company name as the only personalization | Every tool does this, so it reads as automation rather than research |
“[First Name], quick question” subject | Recognized as a template before the email is even opened |
Two or more asks in one email | Splits attention, so the prospect defers the decision and never returns |
Feature list in the body | Describes your product instead of their situation |
Calendar link on email one | Asks for time you haven’t earned yet from someone who doesn’t know you |
Paragraphs longer than the preview pane | Looks like work on a phone screen, gets postponed, then forgotten |
Pitching a Manager on board-level outcomes | Wrong altitude, so the value prop doesn’t map to anything they own |
The skill exists to interrupt that list. Instead of writing each first touch from scratch, or recycling last month’s template, you get the same framework enforced on every single draft.
How to download and install the Claude skill .md file
1. Download the .md file from GitHub
Head to the lemlist Claude Skills Hub, find the Copywriting First Touch card, and click Download. Since it’s plain Markdown, you can open it in any text editor to read through the instructions or tweak them before loading it into Claude.
2. Load the skill into Claude
A prompt tells Claude what to do once. A skill tells Claude how to do it correctly, every time, with persona research steps, subject line formulas, CTA patterns, and sequence logic baked into the file.
Loading it depends on where you work:
- On claude.ai: click the paperclip icon, attach the .md file, and send it with a short instruction like “read this skill file and follow its instructions”
- On Claude Desktop: drag the file straight into the chat window
- In Claude Code: drop it in your .claude/skills/ directory, where Claude discovers it automatically
3. Add the skill to a Claude Project for repeated use
If you’re on Claude Pro or Team, you can upload the file to a Project’s Knowledge section. Every new conversation inside that project loads the framework automatically, so there’s no re-uploading each time you start fresh.
lemlist publishes companion skills in the same library, so you can stack a few together and cover more of the outbound workflow in one place. More on those further down.
Cold email copywriting framework inside the skill
The skill enforces four rules on every draft it produces. Each rule exists because skipping it tends to tank reply rates.
Subject line rules
Subject lines stay short and reference something specific about the prospect. No clickbait, no “[First Name], quick question” defaults. A subject tied to a real detail reads like a human wrote it, because one did.
Four patterns the skill works from:
- Trigger as a question: “New office in Berlin?”
- Their own words, quoted back: “‘cut ramp time in half’”
- Their role plus the outcome: “SDR ramp at Acme”
- Flat observation, no pitch: “saw the v3 launch”
Lowercase is fine. Punctuation is optional. What’s not optional is the specific detail.
Opening line and cold email personalization logic
The first sentence has to reference something unique to the prospect or their company. Generic openers and company-name-only personalization get blocked by design. Feed the prompt richer context, like their LinkedIn summary and a recent company announcement, and the opening line gets noticeably sharper.
Four opening line patterns:
- Event plus consequence: “Saw Acme opened an EMEA office last month. Usually that means the first two AEs are ramping on a playbook that doesn’t exist yet.”
- Their content, reflected: “Your post on outbound attribution said reps spend more time reporting than selling. Heard the same thing from four VPs this quarter.”
- Hiring signal: “You’ve had three SDR roles open since June.”
- Stack change: “Noticed Acme moved off Outreach.”
Notice what none of them do. They don’t compliment the prospect, and they don’t explain who you are. Both can wait.
This is where enrichment data changes the output the most. lemlist’s data enrichment agents pull structured details from LinkedIn, company websites, and internal sources like CRM records and Claap sales recordings, so you can paste that context straight into the prompt instead of researching each prospect by hand.
Body structure and word count constraints
One value proposition per email. One. The skill caps word count and blocks feature lists, multiple asks, and long paragraphs, so the body stays focused on connecting the prospect’s situation to a single clear reason to reply.
CTA patterns and sequence bridge
CTAs stay low-friction: a question, or a permission-based ask. Asking for a meeting on email one hasn’t been earned yet, so the skill avoids it. If this email opens a sequence, the skill can add a brief bridge line hinting at what comes next.
Calibrate the altitude to seniority
Same product, same trigger, different email depending on who’s reading it. This is the rule most reps skip, and it’s the one that quietly decides whether your value prop lands.
Write to a VP about the outcome they’re accountable for in a board meeting: pipeline coverage, ramp time, cost per meeting. Keep it short, because they’re reading on a phone between meetings.
Write to a Manager about the process they run day to day: how their team hits quota, what breaks when two reps leave, what reporting takes off their plate. They’ll forgive a slightly longer email if it’s specific to their team.
Write to an IC about the task they personally hate. Not strategy, not ROI. The twenty minutes of manual work that disappears. Tell the prompt the seniority level and the skill adjusts the value prop and the length accordingly.
How to write your first prompt after loading the skill
Once loaded, the skill stays active for the rest of the conversation. You can ask Claude to adjust tone, add more personalization, change the angle, or process a new batch of leads, and the framework carries through each time.
Start simple and add detail as you go:
- “Write a first-touch cold email for [Name], [Title] at [Company].”
- “Write a first-touch cold email for [Name], VP of Sales at a mid-market SaaS company. Our product helps reduce SDR ramp time.”
- “Here’s their LinkedIn summary: [paste]. Here’s recent company news: [paste]. Write the first-touch email.”
- “Write a first-touch email that opens a 3-email sequence, with a bridge to email two.”
More context in the prompt produces better personalization in the output. If you’re working a large list, lemlist’s enrichment agents can pull verified emails, phone numbers, LinkedIn context, and company news automatically, so you feed the skill real material instead of guesses.
💡 Tip: Paste enrichment data directly into the prompt instead of summarizing the prospect yourself. The skill performs noticeably better with raw details than with your paraphrase of them.
Example first-touch cold email output
Here’s an anonymized draft the skill produced for a VP of Sales at a mid-market SaaS company that had just posted three SDR roles:
Subject: three SDR roles open
Saw you’ve had three SDR openings posted since June. That usually means the first 90 days of ramp are getting rebuilt by whoever has time that week.
We cut new-rep ramp at similar teams by giving each SDR a sequence library that’s already benchmarked, so week one looks like month three.
Worth a look, or is ramp already handled on your end?
And here’s how that draft maps back to the rules that shaped it:
Email element | Framework rule applied |
|---|---|
Subject line | Capped length, references the hiring signal, no template phrasing |
Opening line | Pulled from a real job posting, states the consequence, no filler |
Body | Single value prop tied to the prospect’s situation, under word count |
CTA | Low-friction question, no calendar link |
The output is a starting draft, not a finished email. Read it over, adjust it to sound like you, and double-check any prospect details before hitting send. The skill handles structure. You bring the judgment.
How to customize the skill for your ICP and industry
Edit the .md file to match your offer and persona
Since the file is plain Markdown, opening it in a text editor lets you find the sections defining product description, target persona, and tone. Adjust those once, and the skill applies your version to every draft going forward.
Feed enrichment data into the prompt for deeper personalization
The skill’s output only gets as specific as what you feed it. Swap “they’re a SaaS company” for the actual company description, LinkedIn headline, funding news, or tech stack detail, and the difference shows up immediately in the draft.
lemlist’s enrichment agents pull this context automatically from LinkedIn, company sites, CRM data, and Claap recordings. You can export it and paste it into Claude, or, if you’re using lemlist MCP, skip the copy-paste step entirely and let the agent fetch it.
How the first touch skill connects to follow-up and sequence skills
Email one is only part of the sequence. A typical cold outreach sequence runs two to four emails, and each follow-up needs different rules than the opener.
lemlist publishes companion skills for the rest of that sequence:
- Copywriting Follow Up: writes follow-ups after no reply, never repeating the first email and always introducing a new angle
- Copywriting Analyzer: scores existing cold emails against benchmarks and rewrites the parts that are failing
- Outbound Analyst: benchmarks your campaign stats against data from 244K+ real campaigns and 249M+ emails
Loading multiple skills into one Claude Project means you can move from first touch, to follow-up, to campaign analysis, without leaving the conversation.
Turn first-touch emails into live outbound campaigns
A draft is only useful once it’s running in a live sequence. From here, you have two paths.
The manual path: copy the email into lemlist, build a multichannel sequence across email, LinkedIn, calls, SMS, and WhatsApp from the dashboard, then manage replies across all of those channels in lemlist’s unified inbox. It takes a few minutes per campaign and works fine if you’re launching one or two a week.
The MCP path removes the copy-paste. MCP (Model Context Protocol) is an open standard that lets AI agents call external tools, and lemlist’s MCP server exposes 40+ actions covering lead search, enrichment, sequence building, and campaign management. Setup takes under 2 minutes: paste the server URL and your API key into Claude, Cursor, Windsurf, or your MCP client of choice. From there, Claude can find prospects in lemlist’s 600M+ lead database, enrich them with verified contact details, write the first touch using the skill, and launch the sequence, all without switching tabs.
That’s the difference between using Claude as a copywriter and using it as an outbound operator.
FAQs about the copywriting first touch Claude skill
Can I use the copywriting first touch skill without a lemlist account?
Yes. The skill is a standalone .md file that works with any Claude plan. A lemlist account only comes into play if you want to launch campaigns, enrich leads, or connect through MCP.
Does the first touch skill work with Claude Code or only Claude Desktop?
It works with both. In Claude Desktop, you drag the file into the chat. In Claude Code, custom skills are filesystem-based and need no API upload: place them in ~/.claude/skills/ for personal use or .claude/skills/ for a specific project, and Claude discovers and applies them automatically.
How many cold emails does the skill generate per prompt?
One complete email per prompt, covering subject line, body, and CTA. You can iterate within the same conversation by changing prospect details or asking for a different angle.
What prospect data produces the best results?
At minimum, name, title, and company get the skill started. Adding recent company news, a LinkedIn summary, or an intent signal usually produces a sharper draft, and lemlist’s enrichment agents can pull that data automatically if manual research isn’t an option.
Final thoughts
The skill won’t make a bad offer good. Nothing will. What it does is stop the four failures that show up in nearly every flat sequence: the generic opener, the company-name personalization, the stacked asks, and the calendar link on email one. Fix those and your follow-ups get a real audience instead of an archived thread.
Download the .md file, run five prospects through it, and compare the drafts to what you sent last month. If the output is sharper, the next step is getting it into a sequence that actually sends.
Start a 14-day free trial
</article>
</article>
Share this post