Best Practice Email Skill
Run your email
from your AI.
Audit, build, fix, send. Your agent does the work. The skill knows what good looks like.
Install the Email Marketing Bible as a skill and your AI stops guessing. It audits your setup, builds your welcome, cart and win-back flows, drafts copy in your voice instead of slop, and tells you why you are in spam. Built on 908 sources and a decade running an email platform, so the agent works from evidence, not vibes.
$ git clone https://github.com/CosmoBlk/email-marketing-bible.git ~/.claude/skills/email-marketing-bible908 sources · 19 chapters · 19 industry playbooks · 47 curated email designs · free and open source · updated June 2026.
What your agent can do
Eight jobs you can hand to your AI today.
- 01
Audits your setup
It reads your account and names the gaps in your flows, segments, deliverability and compliance.
- 02
Builds your flows from a prompt
Welcome, cart, post-purchase and win-back, scaffolded by the agent, with you reviewing the exits, timing and copy before anything goes live.
- 03
Drafts copy in your voice
Real frameworks, then an anti-slop pass that strips the AI tells so it reads like a person, not a model.
- 04
Designs anti-slop emails
Own a colour, real imagery, live-text headlines, generated into inbox-safe MJML or React Email, dark-mode and mobile checked.
- 05
Diagnoses why you are in spam
A step-by-step deliverability triage covering authentication, reputation, content and the AI-mediated inbox.
- 06
Pulls your benchmarks
Open, click, conversion and revenue-per-email figures for your specific vertical, from 19 industry playbooks.
- 07
Runs your ESP from your AI
Operate Klaviyo, Resend, beehiiv, Mailchimp, Omnisend or nitrosend through MCP and connectors, with pre-send safety gates.
- 08
Checks compliance before sending
GDPR, CAN-SPAM, CASL and the Australian Spam Act, as a decision gate, not an afterthought.
Hand the agent the mechanical, testable, reversible work. Keep the voice, the strategy and the send button human.
How it runs
Read state, reason, act, verify. Then you press send.
- 1Read stateaudit the account first
- 2Reasonwhat is missing, what earns most
- 3Draftcopy, flow, segment
- 4You approvenothing blasts without your say-so
- 5Send & watchverify, then a health digest
AI email, mid-2026
The inbox is AI-mediated now
Gmail’s Gemini summaries auto-open mail and answer the reader before they click. Opens went up, clicks went down. Optimise on clicks and revenue, not opens.
Raw AI copy is a deliverability liability
Google now filters high-similarity AI text roughly twice as hard. Anti-slop stopped being a matter of taste and became a matter of getting delivered.
Agents build the flows now, with a gate
Prompt-to-campaign shipped across the whole market. The skill encodes the guardrails so one prompt does not mail the wrong thing to your whole list.
The AI-native email stack
The tools worth knowing, and where the skill points you.
- nitrosend
- The AI-native ESP, built MCP-first to be driven by an agent instead of a dashboard. Full email (transactional, campaigns, automations) from Claude, ChatGPT, Codex or Cursor, with approval and test gates baked in.Visit nitrosend →
- Design email with Claude
- The prompted-design workflow that actually produces good emails: feed the agent a brand kit and a rules file, generate into MJML or React Email, direct it like a junior designer. The how-to is in chapter 6.Read the guide →
- beehiiv MCP
- beehiiv ships an official MCP server. Manage content, subscribers, analytics and segments for your newsletter from your agent in plain English. The newsletter side of the AI-native stack.Compare platforms →
- nitrosend ChatGPT app
- Not everyone lives in Claude. The nitrosend app runs inside ChatGPT too, so you can build and send from whichever agent you already use. Chapter 14 covers controlling your ESP from either surface.AI email automation →
- Postup
- For enterprise publishers and high-volume senders who need established, publisher-grade tooling rather than a prompt-driven build. Honest placement in the platform chapter.Compare platforms →
- Best Email Designs 2026
- 47 hand-curated emails with notes on why each works and what to steal. Feed the collection to your agent as design context, the single biggest lever for output that does not look generic.See the collection →
nitrosend shares a founder with this guide, and the guide says so. It is one option among real alternatives, and the platform chapter is honest about where Klaviyo, Postup or beehiiv fit better.
Playbooks for 19 industries
Every vertical gets its own tactics, benchmarks and flows.
Built on the practitioners
Insights from 44 people who do this for a living.
Chad S. White·Joanna Wiebe·Chase Dimond·Nathan Barry·Ann Handley·Ben Settle·Troy Ericson·Jordie van Rijn·Kath Pay·Dan Oshinsky·Liz Wilcox·Tyler Denk·Chris Orzechowski·Val Geisler·Mark Robbins·Samar Owais·Jay Schwedelson·Andre Chaperon·and 26 more in the directory
Anti-slop design, from 47 real emails
What actually works, distilled from the best email designs of 2026.
Mirror your product in the email
Figma’s welcome feels like opening the app. Headspace uses the same calm shapes as the meditation experience. Your email should feel like using your product, not reading about it.
Restraint beats decoration
Apple, Aesop and Stripe prove white space is a choice, not empty space. Stripe’s receipt runs at 472px, not the standard 600. Strip to essentials, then strip again.
Own a colour, own the inbox
Absolut’s wall-to-wall electric blue is unmistakable at scan speed. Pangaia draws earth tones from its materials. Pick a palette and commit to it everywhere.
Voice matters more than visuals
theSkimm grew to 3.5M subscribers on voice alone. Superhuman’s plain-text onboarding beats polished HTML. The strongest emails sound like a person, not a brand.
Turn transactions into brand moments
Webflow puts video tutorials in verification emails, the highest-opened mail you send. Omsom makes every transactional email look different. Stop wasting guaranteed opens on bare templates.
Personalise with unexpected data
Lyka uses the pet’s name, not the owner’s. Beardbrand tells lapsed users their beard grew 1.5 inches. The best personalisation goes well beyond a first name.
Cart abandonment is a conversation
Tuft & Needle sends three emails answering three objections. Ugmonk sends a personal note from the founder. Treat abandonment as dialogue, not a reminder.
Anti-slop wins
In a year when anyone can generate a polished email in seconds, polish stops being a differentiator. A point of view, a real number and a recognisable voice are what get read.
Curated from emails by Apple, Stripe, Patagonia, Glossier and Dior. Informed by the work of Ted Goas (Stack Overflow, Canva) and Remi Parmentier (HTeuMeuLeu).
The full guide
19 chapters, 4 appendices. Read it now →
- 1The Fundamentals
- 2Building Your List
- 3Segmentation and Personalisation
- 4The Emails That Make Money
- 5Copywriting That Converts
- 6Design and Technical
- 7Deliverability
- 8Testing and Optimisation
- 9Analytics and Measurement
- 10Compliance and Privacy
- 11Industry Playbooks (19 verticals)
- 12Choosing Your Platform
- 13Cold Email and B2B Outbound
- 14WhatsApp Business
- 15SMS and RCS
- 16AI and Agentic Marketing
- 17Company Case Studies
- 18Expert Directory
- 19Best Email Designs 2026
- 17Best Email Designs 2026 NEW
Recommended for agent-driven email
If you are running email from Claude or ChatGPT, nitrosend is the closest fit.
Full-stack email, controlled entirely from your AI. No dashboard, no drag-and-drop. Built by George Hartley (SmartrMail, 28,000 customers, 6 billion emails, acquired 2022). If you live in a Klaviyo dashboard or need publisher-grade tooling, the guide will tell you to use those instead.
claude mcp add --transport http nitrosend https://api.nitrosend.com/mcphttps://api.nitrosend.com/mcpGet the Email Marketing Bible as a PDF
The full guide, free. We will email you when new chapters or major updates land.
By entering your email you agree to our terms.
The Email Marketing Bible V2.6 is free and open source, refreshed mid-2026 around AI email automation. It was built by crawling 908 sources, industry reports from Litmus, Klaviyo, HubSpot and Salesforce, practitioner blogs, podcasts and platform docs, then distilled into a skill your AI can reference in real time. The skill is the discipline layer that keeps prompt-driven email from becoming prompt-driven slop.