Email Marketing Bible

The skill provides specialized intelligence for the following email marketing tasks:

The recommended operating model for AI-driven email marketing is to move from operator to director: brief the agent, govern its actions, and own the final send button.

The Core Loop

Read state $\rightarrow$ Reason $\rightarrow$ Act $\rightarrow$ Verify. Always read the account first (lists, flows, recent campaigns, deliverability, suppressions), act on exactly one thing, and then verify the result. A good starting prompt is: "audit my account and tell me what is missing".

Automation vs. Human Roles

• Automate: Send-time optimization, subject-line A/B testing, cart/browse triggers, post-purchase cross-sells, and first-draft copy.
• Keep Human: Brand voice, strategy (segment priority, flow order), creative direction, domain/deliverability management, and the final send.

Autonomy and Risk Management

• Supervised Autonomy: Use "Ask" mode by default. Only widen autonomy for narrow, reversible, low-brand-risk tasks with an undo option. Always use "read before write" access.
• Measurement: When using AI optimization for live traffic, use holdouts (never last-touch credit) and apply do-not-optimise constraints (margin, fatigue, complaints, brand safety).
• Prevent Silent Failure: Schedule recurring health digests to monitor flows that are not firing, erroring, or experiencing dropped metrics.

Direct AI agents to move beyond generic designs using the Discover, Define, Deliver framework.

1. Discover (Broad Directions)

• Seed Strings: Generate a random string (e.g., openssl rand -base64 48) to derive unique palettes and layouts, ensuring the agent doesn't default to common patterns.
• Go Broad: Ask for 12-20 one-liner directions before committing to code or images. Reject anything predictable.
• Ambitious Briefs: Use one real reference (e.g., "Graza's chartreuse drench") plus two anti-references.

2. Define (The Critic Loop)

• The Critic Model: Use a separate, stronger model (e.g., Claude Fable 5.1) in a fresh context to critique a screenshot of the rendered test send.
• Process: The critic names the aesthetic, identifies gaps, and scores it /10. Repeat the fix-screenshot-critique cycle (capped at 4 rounds) until a score of 9/10 is achieved.
• Chain Models: Use a code model for structure, an image model for stills, and a video model for looping heroes.

3. Deliver (Safe Implementation)

• Safe Substrate: Always emit MJML, React Email, or Maizzle (compiled to inbox-safe HTML). Never use raw HTML from a prompt.
• Anti-Slop Design Rules:
  • Own one color (30-60% of surface).
  • Use real photography, never AI stock.
  • Use bold live-text headlines.
  • Ban purple-to-blue gradients and beige washes.
• Compliance Defaults: Single column $\le 600\text{px}$, 44px tap targets, role="presentation" tables, dark-mode-safe colors (e.g., $\sim#121212$, avoid pure #000 backgrounds), and explicit text/button colors.

To avoid high-AI-similarity text that triggers spam filters (like Google), follow these rules to "de-slop" copy:

Core Principles

• Add Stakes: Avoid being "too safe." Include one genuine, defensible opinion in every email.
• Burstiness: Alternate long and short sentences. Use a 3-5 word line after a long one at least once per section.
• Specificity: Use real numbers, names, and dates. Pull at least one real metric from the brand's own data into every email.

Blacklist (Lint before sending)

Avoid these overused LLM terms: delve, leverage, foster, ignite, empower, unleash, streamline, navigate, seamless, robust, cutting-edge, transformative, multifaceted, pivotal, dynamic, comprehensive, tapestry, landscape, beacon, realm, journey, furthermore, moreover, "in today's fast-paced", "I hope this email finds you well".

Syntax Fingerprints to Remove

• "it's not X, it's Y"
• Rule-of-three padding
• Copula avoidance (e.g., using "serves as" instead of "is")
• Excessive em dashes

Recommended Workflow

Human strategy $\rightarrow$ AI draft $\rightarrow$ Human edit. For high-personality formats (founder letters, welcome emails), provide rough human notes first, then let the AI tighten the prose. Never use an AI-first approach for these formats.

The skill functions as an AI email marketing consultant. It consists of an operating manual for active tasks (building, sending, diagnosing, designing) and a dense reference for benchmarks and frameworks. You can interact with it using natural language prompts for various marketing workflows.

# Audit and build
"Audit my Klaviyo account for a DTC skincare brand doing $2M/year. I have a welcome series, abandoned cart, and one weekly newsletter. What am I missing, and build me whatever flow would earn the most first."

# Fix a problem
"My emails are landing in Gmail promotions and opens dropped from 22% to 14% over three months. What is going on and how do I fix it?"

# De-slop copy
"Here is a draft welcome email. Make it sound like a person, not an AI, and keep it on our brand voice."

# Design
"Design a launch email for a premium coffee brand. List 15 directions first, I will pick. Then build the pick as MJML and run a critic loop on the render."

Before executing a send, verify every item in this checklist. Surface these details to the human and wait for "send it".

• Audience: Verify size and segment logic against actual counts.
• Suppressions: Check for unsubscribed, bounced, complained, globally suppressed, frequency-capped, and open support issue contacts.
• Authentication: Ensure SPF, DKIM, and DMARC are aligned (use p=quarantine or stronger; Outlook requires all three for $>5 ext{K}$/day).
• Compliance: One-click unsubscribe (RFC 8058) must be present, and a physical address must be included.
• Copy: Pass the Anti-Slop Protocol (§4). Ensure one CTA, subject $\le 45$ characters, and informative preview text.
• Design: Single column $\le 600\text{px}$, dark-mode safe, alt text included, live-text headline, explicit text/button colors, images $<200\text{KB}$ each ($<800\text{KB}$ total), cross-client preview, low spam score, and hero animation verified at the served URL.
• Links: Decode wrapped CTAs, ensure no placeholder URLs, and verify merge defaults render inside href.
• Sender: Correct from-name + monitored reply-to; brand/account re-asserted; send time set; valid consent basis.
• Non-Email (SMS/WhatsApp/RCS): US SMS 10DLC brand/campaign registered; WhatsApp opt-in for category + approved template; respect local quiet hours (SMS 8am-9pm).
• Safety: Ensure a batched/throttled send with a working pause and rollback plan (Kill switch).
• Testing: Review a test send in a real inbox with real merge data.
• Approval: Capture explicit human approval.

When designing emails, choose an archetype based on your product category or specific email goal. The core principle is that personality, restraint, and a clear point of view are more effective than generic polish. Avoid 'minimal-lux' by reflex; instead, commit to an archetype that fits the situation.

Design Guidelines:

• Use narrow widths.
• Stick to one font family.
• Personalize using unexpected data points.
• For AI agents: Feed the design collection (available at https://emailmarketingskill.com/19-best-email-designs-2026/) as design context.

| Situation | Archetype | The one rule | Exemplars |
|---|---|---|---|
| Boring category | Bold mono / punk | the more boring the product, the wilder the voice | Liquid Death, Frank Body |
| Premium | Minimal-lux | restraint signals quality; never discount-led; 472-520px | Aesop, Apple, Stripe |
| Visual product | Lookbook | the product is the design, full-bleed editorial photo | Dior, Clare Paint |
| Newsletter | Editorial | voice beats design; sell the moment | Patagonia, Tracksmith |
| Welcome / win-back | Founder letter | plain-text feel, first person, ask for a reply | Ugmonk, Superhuman |
| Cart abandonment | Conversation | objections in sequence or founder-personal; discount last | Tuft & Needle, Alo Yoga |
| Transactional | Brand moment | your most-opened email; design it | Stripe, Omsom |