Updated September 28, 2026 | 12 min read
Updated September 28, 2026 | 12 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.
Copywriting Refiner Claude Skill: I Ran It Against 8 Rules on Every Cold Email (Here's What Failed)
Last quarter, I ran our own outbound sequences through a copy audit. The results were bad. Two out of every three first-touch emails failed at least one basic quality check: bloated word counts, “I” openers, early meeting asks, generic flattery dressed up as personalization. And these were emails written by people who do this for a living.
That experience is why I keep coming back to the Copywriting Refiner. It catches the mistakes you stop noticing after you’ve stared at the same draft for twenty minutes.
The Copywriting Refiner is a Claude Skill built by lemlist that audits cold emails and LinkedIn messages against eight fixed quality checks, then rewrites whatever fails. Paste in a message, ask for an audit, and you get a pass/fail report plus a corrected version. Every time, no matter how you phrase the request, whether you say “refine this email,” “clean up my copy,” or just “audit this” (ColdIQ Skills Directory).
A Claude Skill is an organized folder of instructions, scripts, and resources that agents can discover and load dynamically to perform better at specific tasks. (Equipping agents for the real world with Agent Skills \ Anthropic) Anthropic published the specification as an open standard (see Anthropic’s Agent Skills documentation). In practice, think of it as a playbook you hand to Claude before the conversation starts. Once the skill is loaded, Claude follows those instructions instead of improvising, which matters when the task is “review my copy the same way every time.”
That consistency is the actual value. Instead of vague feedback that shifts depending on how you word your prompt, you get the same eight checks and the same output format, whether it’s your first email this week or your fiftieth.
Eight quality checks the Copywriting Refiner runs on every message
So what exactly does the skill look for? Every message gets scored against the same eight checks, pass or fail, and each failure comes with the exact phrase that triggered it. That last part matters: you’re not left guessing what “too salesy” means in your specific email.
Six of the checks apply across formats.
The skill flags any em dashes or en dashes on sight. Dashes make an email read like a brochure instead of a message from one person to another. Flagged instances get swapped for commas, periods, or parentheses.
Rhetorical questions used as openers also fail. “Are you tired of X?” is one of the most overused hooks in cold outreach. If the question doesn’t genuinely need an answer from the prospect, it gets cut.
Generic flattery like “Love what you’re building” fails immediately. A compliment needs to reference something verifiable, like a funding round or a specific post, or it reads as manipulation.
Pompous jargon gets the same treatment. Words like “synergies” and “best-in-class” have been used so often they’ve stopped meaning anything. The skill replaces them with a plain description of what the product actually does.
Starting the body with “I” fails because it makes the message about the sender. The rewrite leads with the prospect’s context instead.
Requesting a call in the first email, before the prospect has agreed the problem is worth solving, also fails (unless you’ve explicitly told the skill you want that CTA).
The other two checks are more mechanical, but just as easy to get wrong. Message length has hard limits by format:
Format | Word limit |
|---|---|
Email 1 (first touch) | 120 words |
Email 2+ (follow-up) | 150 words |
LinkedIn Message 1 | 60 words |
LinkedIn Message 2 | 80 words |
Lemlist’s campaign data shows that 300+ word emails have only a 19% booking rate, compared to 52% for emails around 120 words. (How Long Should a Cold Email Be? (2026 Data) | Overloop) That’s why the first-touch ceiling sits at 120 words, not some arbitrary number. The LinkedIn limits are tighter for the same reason: shorter messages outperform on that platform by an even wider margin.
Subject lines get their own rule: six words max, sentence case, no exclamation marks, and no filler like “Quick question.” It’s a small detail, but it’s often the difference between an email that gets opened and one that doesn’t.
How to install the Copywriting Refiner
Getting the skill running takes about two minutes.
Installing via Claude.ai (no code)
- Download the SKILL.md file. Go to lemlist’s Claude Skills page and grab the Copywriting Refiner file. You can also find it in the skill’s open-source repo on GitHub.
- Load it into Claude. Open a Claude Project and upload the file to the project’s knowledge section. Every new conversation in that project will have the skill active from that point on. Since February 2026, Projects are available on Claude’s free tier (Is Claude Free? Claude Free vs Paid Plans Compared | AI Assistants | Gmelius) (Is Claude Free? Claude Free vs Paid Plans Compared | AI Assistants | Gmelius), so you don’t need a paid plan for this. Pro or Team plans give you more usage headroom if you’re auditing at volume.
- Verify it worked. Paste any cold email and type “audit this.” A correctly loaded skill responds with a structured pass/fail table and a rewrite underneath it. If you get a generic paragraph of suggestions instead, double-check that the file is in project knowledge, not just pasted into the chat.
- Run it on a second message type to confirm. Try a LinkedIn DM after the email. The skill should automatically detect the format and apply LinkedIn-specific limits (60 words for a first message, no subject line check). If it applies email rules to a LinkedIn message, re-upload the file.
Installing via Claude Code
If you use Claude Code, you can install the skill directly from the command line:
npx github:l3mpire/claude-skills copywriting-refiner
Restart Claude Code (or start a new session) and the skills become invokable (GitHub - l3mpire/claude-skills · GitHub), since Claude will discover them automatically based on the skill’s description field. For more on how lemlist skills work inside Claude, see the step-by-step guide to using Claude Skills.
How to run your first outreach copy audit
Once the skill is loaded, the workflow itself is short.
Start by pasting your full cold email or LinkedIn message into Claude, subject line included if it’s an email. You can just paste and hit enter, or type something like “audit this cold email.” The skill detects the format automatically and applies the matching rules.
Claude then returns a table with all eight checks marked pass or fail, each failure quoting the exact phrase that caused it. A short summary underneath explains what’s dragging the copy down overall. This is where you actually learn something, since seeing your own phrase flagged tends to stick better than a general tip ever does.
Right below the audit sits a rewritten version with only the failing elements fixed. If your value prop was already clear, it stays untouched.
Tip: Read the original and rewritten versions side by side rather than just skimming the changelog. That comparison is what builds your instinct for what strong outreach copy sounds like, faster than reading rules ever will.
Once you’re happy with the rewrite, copy it into your outbound tool. lemlist users can paste directly into a sequence step, and teams on lemlist MCP can push the updated copy into a live campaign from inside Claude without switching tabs.
Before and after examples of audited cold emails
Seeing the output tells you more than reading about it. Here are three common patterns the skill catches.
Example 1: First-touch email (four failures)
Element | Before | After |
|---|---|---|
Subject line | “Quick Question About Your Growth” | “your SDR ramp time” |
Opening line | “I came across your company and love what you’re building.” | “Noticed you posted 2 SDR roles last week.” |
CTA | “Would you be open to a 15-min call?” | “Want the breakdown we used with [similar company]?” |
This one fails four checks at once: generic flattery, an “I” opener, an early meeting ask, and a Title Case subject line over six words. The rewrite keeps the original angle and grounds each line in something the prospect can verify, while swapping the meeting ask for a lower-friction offer.
Example 2: LinkedIn first message (two failures)
Element | Before | After |
|---|---|---|
Opening | “I saw your profile and thought we could synergize on scaling outbound.” | “Your team doubled headcount in Q2, curious how outbound is keeping up.” |
CTA | “Let’s hop on a quick call this week?” | “Happy to share the playbook if useful.” |
Word count | 74 words | 58 words |
The skill flags “synergize” as pompous jargon and catches the early meeting ask. It also trims the message to fit within the 60-word LinkedIn first-message limit.
Example 3: Follow-up email (passes all checks)
Element | Before | After |
|---|---|---|
Subject line | “re: outbound” | No change needed |
Opening | “Circling back on the ramp-time data I mentioned.” | No change needed |
CTA | “Want me to send the case study?” | No change needed |
Word count | 43 words | 43 words |
Follow-ups play by different rules. Meeting asks are allowed after the first touch, word limits are more generous (150), and the tone can be more direct. This message passes all eight checks. I include this example because the skill isn’t designed to rewrite everything. When copy is clean, it says so.
How the Copywriting Refiner fits into your outbound workflow
The Copywriting Refiner is a review step. It sits between writing a first draft and loading the message into your campaign.
- Draft. Write or generate the initial copy. You can use Claude with one of lemlist’s drafting skills (like the Copywriting First Touch skill), write in a doc, or use lemlist AI to generate your first version.
- Audit. Run the Copywriting Refiner to flag weak spots.
- Fix. Accept the rewrite, or adjust it manually based on the audit.
- Launch. Push the final version into your sequence.
I ran this loop on our own sequences before shipping the skill externally. The biggest surprise was how many follow-ups were just rephrased versions of email one, with the same CTA and the same word count problem. The skill caught it because each follow-up gets checked against its own rules (longer limits, meeting asks allowed) rather than the Email 1 standard. The result is a sequence that builds on itself instead of repeating the same pitch four times with different subject lines.
Copywriting Refiner vs. generic AI copywriting prompts
You might be wondering why this beats just asking ChatGPT or Claude to “improve my cold email.” You can do that, but the feedback changes depending on your prompt wording and how the model happens to interpret “better” that day.
There’s also a category of dedicated email-grading tools worth knowing about. Lavender AI is a real-time email coaching tool that lives inside Gmail and Outlook as a Chrome extension. It scores every cold email you write from 0 to 100 as you compose it. (Lavender AI Review 2026: Is It Still the Best AI Email Coach?) Lavender’s pricing in 2026 starts at a free tier (5 emails/month), with paid plans from $29/month up to $69/user/month for teams. (Lavender AI Review 2026: Is It Still the Best AI Email Coach?) It’s a good product for in-inbox coaching.
The Copywriting Refiner solves a different problem. It’s a free, open-source markdown file you can read and edit in any text editor. There’s no separate subscription, no browser extension to install, and it runs inside the same tool you already use to draft and push copy. You can add checks, remove them, or rewrite the banned-jargon list to match your team’s standards. Try doing that with a SaaS grading tool.
Here’s how the approaches compare side by side:
Dimension | Generic AI prompt | Dedicated email grader (e.g., Lavender) | Copywriting Refiner skill |
|---|---|---|---|
Evaluation criteria | Undefined, changes each time | Proprietary scoring model | Fixed eight-check list, fully editable |
Output format | Freeform paragraph | 0–100 score with suggestions | Pass/fail table plus rewrite |
Repeatability | Varies by prompt wording | Consistent within the tool | Consistent across every audit |
Channel awareness | Treats email and LinkedIn the same | Email-focused | Adjusts limits per format |
Feedback specificity | “Consider shortening your email” | Highlights sections to improve | “FAIL, 147 words, limit is 120” |
Cost | Free (with any LLM) | $29–$69/mo per user | Free, open-source |
Customizability | Depends on prompt engineering | Limited to vendor settings | Edit the markdown file directly |
The practical upside shows up at the team level. When every rep runs their copy through the same eight checks before sending, the floor for outreach quality rises across the whole team, not just for the person who happens to write good prompts.
Start auditing outreach copy before your next campaign
Every cold email that goes out unreviewed is a small bet that it doesn’t have an “I” opener, an early meeting ask, or a subject line that reads like a template. The Copywriting Refiner removes that bet before you hit send.
Download the skill from lemlist’s Claude Skills page, load it into Claude, and run it on your next message before it reaches an inbox. lemlist is rated 4.6/5 on G2 from 1,300+ reviews. Looking to launch the campaigns themselves too? Start a 14-day free trial of lemlist, no credit card required.
Frequently asked questions about the Copywriting Refiner
Does the Copywriting Refiner work with LinkedIn messages, or only email?
Both. It detects the format automatically and adjusts the rules, applying different word limits for LinkedIn and skipping the subject line check entirely, since DMs don’t have one.
Can I edit the checklist to match my team’s standards?
Yes. The skill is a plain markdown file, so opening it in any text editor lets you add checks, remove them, or update the banned jargon list. Claude follows whatever’s written into the file.
How is this different from lemlist’s sequence-writing skills?
Sequence-writing skills like the Copywriting First Touch skill generate new campaign copy from a blank page. The Copywriting Refiner works in the opposite direction, taking copy you’ve already written and checking it against a fixed standard.
What Claude plan does this require?
Since February 2026, Anthropic opened up Projects to free users (Is Claude Free? Claude Free vs Paid Plans Compared | AI Assistants | Gmelius) (Is Claude Free? Claude Free vs Paid Plans Compared | AI Assistants | Gmelius), so the skill works on any Claude plan. You can load it into a Project and keep it active across conversations. Pro or Team plans give you more usage and longer context windows, which helps if you’re auditing a full multi-step sequence in one conversation.
Does the Copywriting Refiner cost anything?
No. The skill itself is a free, open-source markdown file hosted on lemlist’s public GitHub repo. You can download it, edit it, and use it without paying anything beyond whatever Claude plan you’re already on (including the free tier).
Can I use this in Claude Code instead of Claude.ai?
Yes. Run
npx github:l3mpire/claude-skills copywriting-refiner in your terminal to install the skill directly into Claude Code. Restart Claude Code or start a new session and the skills become invokable (GitHub - l3mpire/claude-skills · GitHub) automatically. This path is useful for developers who prefer the CLI over the web interface.
Share this post