How to write your own skill

Writing a skill means encoding how you — or an expert you trust — actually think through a specific marketing problem. The best skills aren't prompts. They're compressed expertise.

What makes a skill good

Before you write a line, ask yourself: what does an expert do that a generalist doesn't?

A generalist asks "what should I write?" An email specialist asks:

• What's the goal of this email in the sequence?
• What does this segment know and believe right now?
• What's the one action I want them to take?
• What objection will they have, and how do I address it?
• What subject line framework fits this email type?

That thinking process — those questions, in that order — is what belongs in a skill.

The SKILL.md format

A SKILL.md file has a few key sections:

1. Role definition

Tell the AI who it's becoming. Be specific.

## Role

You are a senior B2B email marketing strategist with 10+ years of experience
building email programs for SaaS companies. You think in sequences, not
individual emails. You optimize for reply rate, not open rate, for cold outreach
— and for click-to-conversion for nurture.

2. Core frameworks

List the mental models the expert uses. These are the things that make the output different.

## Frameworks

**The 3-part subject line formula:** Specificity + Relevance + Curiosity gap
- Weak: "Improve your email marketing"
- Strong: "How [Company] cut churn 23% with one email sequence"

**Sequence logic:** Every email in a sequence has a job.
Welcome → Educate → Activate → Convert → Retain.
Map each email to a stage before writing it.

3. Decision trees

Help the AI make the right choices without you having to specify.

## When to use plain text vs HTML

Use plain text when:
- Sending to cold prospects (feels personal, not promotional)
- Re-engagement campaigns (stands out in inbox)
- C-suite audiences (they don't read newsletters)

Use HTML when:
- Sending to engaged subscribers
- Announcing a product or feature
- Weekly newsletters with consistent branding

4. Output format

Define what the AI should actually produce.

## Output format

For each email, provide:
1. Subject line (3 options, with framework label)
2. Preview text
3. Full email body
4. Internal note: what job this email is doing and why this approach

5. Quality bar

Tell the AI what good looks like — and what to avoid.

## Quality bar

Good: Specific, concrete, shows understanding of the reader's context.
Bad: Generic advice that could apply to any company.
Never: Clichés ("excited to share", "hope this finds you well", "synergy").

Keep it short

The best skills are focused. Resist the urge to put everything in one file. A skill that covers email marketing comprehensively is less useful than five skills that each cover one type of email deeply.

Target 200–500 lines. If it's longer, split it.

Testing your skill

Before publishing, test it:

1. Install the skill in your AI tool
2. Run 5–10 real tasks you'd actually want help with
3. Compare output to what you'd get without the skill
4. Refine until the difference is obvious

If you can't tell the difference, the skill isn't doing enough work.

Publishing to the directory

To add your skill to mktgskills.ai:

1. Push your SKILL.md to a public GitHub repo
2. Open an issue on the mktgskills repo using the Submit Skill template
3. Include: skill name, repo URL, category, one-sentence description
4. We'll review and add it within a few days

The best submissions include a real example of input → output that shows what the skill does differently.