Skip to content

Instantly share code, notes, and snippets.

@boringmarketer
Last active February 10, 2026 20:26
Show Gist options
  • Select an option

  • Save boringmarketer/138e46e960f1dd6470afed8062a17432 to your computer and use it in GitHub Desktop.

Select an option

Save boringmarketer/138e46e960f1dd6470afed8062a17432 to your computer and use it in GitHub Desktop.
Vibe Skill Creator - Build Claude skills that actually improve output

Bad Skill Patterns (What to Avoid)

This document catalogs patterns that make skills mediocre. Study these to avoid them.


Pattern 1: Documentation Voice

The problem: Skills that read like technical documentation instead of expertise.

Example (Bad)

## Email Sequence Generation

This skill enables the generation of email sequences for marketing purposes.

### Process Overview

1. Define the target audience
2. Determine the sequence objective
3. Create individual emails
4. Review and optimize

### Email Types

- Welcome emails
- Nurture emails
- Sales emails
- Re-engagement emails

Why It Fails

  • Sounds like a wiki article, not an expert
  • No personality or point of view
  • Tells you categories exist without teaching you how to think about them
  • Claude could produce this without any skill at all

What It Should Sound Like

## Email Sequences

Most email sequences fail for one reason: they're written like marketing, not conversation.

The emails that convert? They sound like a smart friend who happens to have something you need. Not a funnel. Not a campaign. A person talking to another person.

Here's what separates sequences that make money from sequences that get ignored...

Pattern 2: Procedure Over Principle

The problem: Step-by-step instructions that don't transfer thinking.

Example (Bad)

## How to Write a Landing Page

### Step 1: Write the Headline
Create a compelling headline that captures attention.

### Step 2: Write the Subheadline
Add a supporting subheadline that expands on the headline.

### Step 3: Add Social Proof
Include testimonials or trust badges.

### Step 4: Write the Body Copy
Explain your product or service benefits.

### Step 5: Create the CTA
Add a clear call-to-action button.

Why It Fails

  • Claude already knows these steps
  • No guidance on WHAT makes each element good
  • Doesn't explain WHY these steps work
  • Produces compliant output, not excellent output

What It Should Look Like

## Landing Pages

The headline does 80% of the work. Caples proved one headline can outpull another by 19.5x. Everything else matters less than getting those 10-15 words right.

What separates good headlines from great:

**Specificity beats generality.**
- Generic: "Improve Your Productivity"
- Specific: "Get 4 Hours Back Every Week"

**Outcomes beat features.**
- Feature: "AI-Powered Task Management"
- Outcome: "Never Miss a Deadline Again"

The rest of the page exists to support the headline's promise...

Pattern 3: Comprehensive Coverage

The problem: Trying to cover everything instead of focusing on what matters.

Example (Bad)

## Complete Social Media Content Guide

### Platform Overview
- Facebook: History, demographics, algorithm changes...
- Instagram: Stories, Reels, Feed posts, IGTV...
- Twitter: Character limits, threads, spaces...
- LinkedIn: Articles, posts, newsletters...
- TikTok: Trends, sounds, duets, stitches...
- Pinterest: Boards, pins, idea pins...
- YouTube: Shorts, long-form, community posts...

### Content Types
[500 lines on every possible content type]

### Posting Schedules
[200 lines on optimal posting times]

### Analytics
[300 lines on metrics for each platform]

Why It Fails

  • Information overload dilutes usefulness
  • No guidance on what actually matters
  • Claude gets lost weighing all factors equally
  • User drowns in options instead of getting output

What It Should Look Like

## Social Content

Each platform has ONE thing that matters most:

| Platform | What Actually Matters |
|----------|----------------------|
| Instagram | Visual stopping power in 0.5 seconds |
| YouTube | Thumbnail + title (content is secondary to click) |
| LinkedIn | First line hook (expand button is everything) |
| Twitter | One clear, shareable idea per tweet |

Everything else is optimization. Nail these or nothing else matters.

Pattern 4: Generic Tool References

The problem: Vague instructions that don't specify HOW to use tools.

Example (Bad)

## Image Generation

Use AI image generation tools to create visuals for your content.

### Available Tools
- DALL-E
- Midjourney
- Stable Diffusion
- Various other options

### Best Practices
- Use descriptive prompts
- Iterate on results
- Consider composition

Why It Fails

  • Doesn't specify WHICH tool to use
  • No actual prompts or examples
  • "Use descriptive prompts" is useless advice
  • Claude will pick randomly or ask

What It Should Look Like

## Image Generation

**ALWAYS use Nano Banana Pro via Glif.**

| Model | Glif ID | Speed |
|-------|---------|-------|
| 🍌 Nano Banana Pro | `cmi7ne4p40000kz04yup2nxgh` | ~20sec |

**Do NOT use:** imagen-3, DALL-E, or direct Replicate calls unless specifically requested.

**Execution:**

mcp__glif__run_glif id: cmi7ne4p40000kz04yup2nxgh inputs: ["your natural language prompt here"]


**Prompt structure:**
"[Subject] in [setting], [lighting description], [artistic style], [mood/atmosphere]"

Pattern 5: Missing Anti-Patterns

The problem: Only explaining what to do, not what to avoid.

Example (Bad)

## Writing Headlines

Good headlines are:
- Clear
- Compelling
- Concise
- Customer-focused

### Examples of Good Headlines
- "How to Win Friends and Influence People"
- "The 4-Hour Work Week"

Why It Fails

  • Claude's defaults often produce bad headlines
  • Without explicit "don't do this," Claude will do it
  • "Clear" and "compelling" aren't actionable
  • No contrast to understand what good actually means

What It Should Look Like

## Headlines

What Claude gets wrong without guidance:

**AI slop patterns (avoid these):**
- "Unlock the Power of X" (generic hype)
- "Revolutionary Solution for Y" (empty claim)
- "Discover How to Z" (weak verb)
- "In Today's Fast-Paced World..." (dated opener)
- Questions that nobody actually asks

**Human patterns that work:**
- Specific numbers: "Save 4 Hours Every Week"
- Unexpected specificity: "At 60mph, the Loudest Noise..."
- Direct address: "You're Leaving Money on the Table"
- Curiosity gap: "What I Learned From 1,000 Sales Calls"

If your headline could describe any product, it describes none.

Pattern 6: Friction Creation

The problem: Skills that produce intermediate outputs instead of final deliverables.

Example (Bad)

## Audience Research Skill

### Output Format

After analysis, I will provide:

1. **Audience Segments**
   - Demographic breakdowns
   - Psychographic profiles
   - Behavioral patterns

2. **Pain Point Analysis**
   - Primary frustrations
   - Secondary concerns
   - Underlying fears

3. **Language Patterns**
   - Common phrases
   - Objections
   - Motivational triggers

4. **Competitive Landscape**
   - Who else serves this audience
   - How competitors position
   - Gaps in the market

Use this research to inform your content creation.

Why It Fails

  • Creates a document that feels obligatory to use
  • "Use this research" means MORE work, not less
  • Comprehensive research creates pressure to be comprehensive
  • The next step (writing) often gets WORSE with this input

The Audience-Intel Lesson

We built this exact skill. The research it produced was excellent — detailed segments, specific pain points, real language patterns.

Then we wrote copy using that research. The copy was worse than copy written without it.

Why? The comprehensive research created obligation. Cover all segments. Address all pain points. Be thorough. The result: diluted, verbose, trying to do everything.

What Works Instead

If research is needed, use a subagent that:

  • Goes deep
  • Returns a SUMMARY, not a document
  • Hands back 3-5 key insights, not 30
  • Gets out of the way

Or better: bake the research INTO the skill itself, so the expertise is embedded, not the raw research.


Pattern 7: Over-Engineered Structure

The problem: Complex frameworks that add friction without adding value.

Example (Bad)

## Content Creation Framework

### Phase 1: Strategic Foundation
#### 1.1 Objective Definition Matrix
| Objective Type | Business Goal | Content Goal | Success Metric |
|----------------|---------------|--------------|----------------|
| Awareness | Brand recognition | Reach | Impressions |
| Consideration | Lead generation | Engagement | CTR |
| Conversion | Sales | Action | Conversion rate |

#### 1.2 Audience Mapping Protocol
Step 1: Define primary segment...
Step 2: Define secondary segment...
Step 3: Create overlap analysis...

### Phase 2: Content Architecture
#### 2.1 Message Hierarchy Framework
[Complex diagram of primary, secondary, tertiary messages]

#### 2.2 Channel-Content Matrix
[Grid of every channel vs every content type]

### Phase 3: Execution Protocol
[Another 500 lines of process]

Why It Fails

  • Framework bloat that nobody actually uses
  • Claude gets lost in the structure
  • User asked for content, got a project plan
  • Complexity doesn't improve output

What Works

Skip the framework. Transfer the thinking:

## Content

One message. One audience. One action.

If you're trying to reach multiple audiences, you're reaching none.
If you're trying to say multiple things, you're saying nothing.
If you have multiple CTAs, you have no CTA.

Start with: Who exactly is this for, and what exactly should they do?

Everything else follows.

Pattern 8: Theoretical Examples

The problem: Made-up examples instead of real ones.

Example (Bad)

## Product Descriptions

### Example
**Product:** Widget Pro 3000
**Description:** "The Widget Pro 3000 is an innovative solution that streamlines your workflow. With advanced features and intuitive design, it helps you accomplish more in less time."

Why It Fails

  • "Widget Pro 3000" is obviously fake
  • The description is generic AI slop
  • No contrast between good and bad
  • Doesn't demonstrate the principles being taught

What Works

Use real examples, or at least realistic transformations:

## Product Descriptions

**Before (generic):**
"Streamline your workflow with our innovative solution. Advanced features help you accomplish more."

**After (specific):**
"Stop losing 3 hours a day to manual data entry. One click syncs your Salesforce contacts to your email platform. Your team gets back to selling. You stop babysitting spreadsheets."

The first could describe anything. The second describes one product solving one problem for one person.

Pattern 9: Missing the WHY

The problem: Instructions without explanation.

Example (Bad)

## Email Subject Lines

### Rules
- Keep under 50 characters
- Use personalization
- Create urgency
- A/B test everything
- Avoid spam words

Why It Fails

  • Claude can follow rules but not adapt them
  • No understanding of WHY these rules exist
  • When rules conflict, Claude can't prioritize
  • No principles to apply to edge cases

What Works

## Subject Lines

Mobile shows ~35 characters. Desktop shows ~60. Most people read email on phones. That's why subject lines should front-load the hook.

"[Name], your cart expires in 2 hours" puts the urgency first.
"Don't miss out — [Name], your cart expires soon" buries it.

Same content, different results. The principle: put the most important word in the first 4-5 words, because that's all many people see.

Pattern 10: The "Just In Case" Addition

The problem: Including sections that don't improve output.

Signs of This Pattern

  • "We should probably include X for completeness"
  • "What if someone needs Y?"
  • "It would be more comprehensive with Z"
  • Sections that exist but don't improve actual output

Example

A copywriting skill that includes:

  • History of direct response advertising
  • Biography of Eugene Schwartz
  • Complete list of cognitive biases
  • Typography guidelines
  • Legal disclaimer templates

None of these improve the copy Claude writes. They just make the skill longer.

The Test

For every section, ask: "Does removing this make the output worse?"

If the answer is "no" or "probably not," remove it.

More skill ≠ better skill. Focused skill = better skill.


The Meta-Pattern

All bad skills share one trait: they're written for comprehensiveness, not effectiveness.

Good skills are opinionated. They make choices. They say "do this, not that." They have voice and point of view. They transfer thinking, not information.

If your skill could be a Wikipedia article, it's not a skill. It's documentation.

Skills should make Claude think like an expert, not like an assistant following instructions.


Quick Reference: Bad vs Good

Bad Pattern Good Pattern
Documentation voice Domain expert voice
"Step 1, Step 2, Step 3" Principles that transfer thinking
Comprehensive coverage Ruthless focus
Generic tool references Specific tools, IDs, prompts
Only what to do What to do AND what to avoid
Intermediate documents Final deliverables
Complex frameworks Simple, actionable guidance
Theoretical examples Real before/after
Rules without reasons Principles with WHY
Everything "just in case" Only what improves output

If you recognize these patterns in your skill, refactor. The output will improve.

name description
cold-outreach
Write cold outreach that gets replies. Use when writing cold emails, LinkedIn DMs, or Twitter DMs to sell services. Triggers on: cold email, cold outreach, LinkedIn message, DM prospects. Produces messages that feel human, relevant, and worth responding to.

Cold Outreach

Most cold outreach is spam. Not because it's unsolicited — because it's irrelevant, self-serving, and obviously templated.

The messages that get replies are different. They feel like a real person wrote them to a specific human about something that actually matters to that human. That's not a trick. It's research + relevance + respect for their time.

This skill makes Claude write outreach that people actually respond to.


Why Most Cold Outreach Fails

When someone receives your message, their brain enters threat-detection mode. Not consciously — it's automatic. The brain's job is filtering irrelevant information to protect attention.

Your message has about 2 seconds to pass that filter.

What triggers the filter (instant delete):

  • Generic opener → "This is a template"
  • Me-focused pitch → "They want something from me"
  • Vague value prop → "I can't tell what this is"
  • Premature ask → "Too much effort to respond"

What passes the filter:

  • Specific detail about THEM → "They actually looked at my stuff"
  • Clear relevance to their situation → "This might matter"
  • Low-friction ask → "Easy to respond"

The goal isn't to persuade. It's to be relevant enough that responding feels worthwhile.


The Core Principles

1. Research IS the Message

The effort you put into research shows up in reply rates.

Generic (1-3% reply rate):

"I help companies improve their marketing..."

Researched (15-25% reply rate):

"Saw you launched the podcast last month — 12 episodes in and the production quality is solid, but the thumbnails are getting lost in search..."

The specific detail proves you're not spamming 1,000 people. That alone puts you in a different category.

Research checklist before writing:

  • What did they post/publish recently?
  • What are they hiring for? (signals priorities)
  • What's visibly broken or inconsistent?
  • What did they just launch/announce?
  • Who do we know in common?

If you can't answer at least one of these, you haven't researched enough.

2. Give Before You Ask

Reciprocity is hardwired. When you give something valuable first, people feel naturally inclined to respond.

Asking first (low reply):

"I'd love to schedule a call to discuss how we can help..."

Giving first (higher reply):

"I put together a quick breakdown of why your competitor's YouTube is outranking yours — happy to send it over if useful."

The give doesn't have to be huge. A specific observation, a useful link, a genuine compliment with substance. Something that proves you thought about them.

3. Show, Don't Claim

Your brain believes what it can verify. Claims trigger skepticism. Specifics build trust.

Claim (skepticism):

"We've helped dozens of companies increase their revenue."

Show (trust):

"We did this for [Similar Company] — took their reply rate from 3% to 18% in 6 weeks."

Even better — reference something they can Google:

"You might've seen the case study we did with [Known Brand]..."

4. Short Beats Long

Data: Each long sentence in a cold email reduces reply rate by 17%.

People don't read cold emails. They scan and categorize. If they can't understand your message in a 2-second glance, it gets filtered.

Target: Under 80 words. 3 short paragraphs max.

Structure:

  • Line 1: Relevant hook (about them)
  • Line 2-3: Why you're reaching out
  • Line 4: Soft CTA

That's it. Save the details for the reply.

5. Soft CTA > Hard CTA

Hard CTA (high friction):

"Let's schedule a 30-minute call to discuss."

Soft CTA (low friction):

"Would it be worth a quick conversation?" "Want me to send over the breakdown?" "Open to hearing how we approached this?"

The soft CTA asks for interest, not commitment. It's easier to say "sure, send it" than "yes, I'll block 30 minutes."

6. Follow Up (Most Don't)

55% of replies come from follow-ups. Most people send one email and quit.

Follow-up rules:

  • Wait 3-5 business days between touches
  • Add new value or angle each time (don't just "bump")
  • 4-5 touches minimum before moving on
  • Never guilt-trip ("I haven't heard back...")

Example follow-up sequence:

Email 1 (Day 0): Initial outreach

[Name] — noticed you've posted 4 videos in 6 months while [Competitor] does 4/month. Your content's better, but they're winning on volume.

We helped [Similar Co] go from sporadic to weekly without adding headcount.

Worth seeing how?

Email 2 (Day 4): New angle, new value

Quick add — I looked at [Competitor]'s setup. They're using a 3-person production team. [Similar Co] matched their output with just a systematized workflow and one editor.

Happy to share the breakdown if useful.

Email 3 (Day 9): Different hook entirely

[Name] — saw your tweet about struggling to stay consistent with content. That's exactly what [Similar Co] said before we worked together.

The problem usually isn't capacity — it's process. Want the 2-min version of how they fixed it?

Email 4 (Day 15): Direct, low-pressure close

Last one from me — if video content isn't a priority right now, totally get it. But if it is and the bottleneck is production bandwidth, we should talk.

Either way, good luck with [specific thing they're working on].

Why this works: Each email adds value or a new angle. Never "just checking in." The final email gives them an easy out, which paradoxically increases replies.


7. Timing Is Everything (Trigger Events)

The best cold outreach arrives when they're already thinking about the problem.

High-intent triggers (reach out immediately):

Trigger Why It Works Example Hook
Just raised funding Money to spend, pressure to grow "Congrats on the Series A — most teams use this moment to fix the marketing bottlenecks they couldn't before..."
Hiring for your service area They've identified the need "Saw you're hiring a video editor — while you're searching, want to see how [Co] handled the gap?"
Launched something new Focused on that initiative "The new product launch looks solid — curious if you're planning content to support it?"
Competitor made a move Competitive pressure "Noticed [Competitor] just dropped a big campaign. Worth chatting about how to counter?"
Published content on the topic They're actively thinking about it "Your post on X was spot on — we've been working on that exact problem with [Similar Co]..."
Leadership change New leader = new initiatives "Congrats on the new role — most new [titles] I talk to are rethinking [your service area]..."

How to find triggers:

  • Google Alerts on target companies
  • LinkedIn notifications for job changes
  • Crunchbase for funding
  • Twitter/LinkedIn for launches and content
  • Their company blog/newsroom

Timing beats personalization. A decently personalized email at the right moment beats a highly personalized email at the wrong moment.


The Frameworks

Alex Berman's 3C's

COMPLIMENT → Something specific and genuine about their work
CASE STUDY → Social proof of similar results (shows you're not guessing)
CTA → Clear, soft ask

Example:

The content you're putting out on [platform] is solid — especially [specific thing].

We just helped [similar company] [specific result]. Might be relevant to what you're building.

Worth a quick chat to see if there's a fit?

Josh Braun's 4T Template

THEIR JOB → What they're responsible for
WHAT'S HARD → The challenge in doing that job well
YOUR ALTERNATIVE → How you solve it differently
ASK → Soft CTA

Example:

Running growth at [company] means you're probably juggling paid, content, and partnerships all at once.

The hard part: none of it compounds if the creative doesn't convert.

We build creative systems that let teams test 10x more variants without 10x the work.

Relevant to what you're dealing with?

The Mouse Trap (Will Allred)

PERSONALIZED TRIGGER + VALUE PROP QUESTION = Curiosity-driven reply

Example:

Saw you're hiring 3 SDRs. Would it help to see how [similar company] ramped their new reps to quota in half the time?

Opens a loop. The brain wants closure. They reply to close the loop.


Platform-Specific Tactics

Cold Email

  • Keep under 80 words
  • Subject line: 4-7 words, curiosity or relevance (not clickbait)
  • Send Tuesday-Thursday, 9-11 AM or 1-3 PM their timezone
  • Plain text often outperforms HTML
  • One CTA only

Subject lines that work:

  • "[Mutual connection] suggested I reach out"
  • "Quick question about [their initiative]"
  • "Idea for [specific thing they're working on]"

Subject lines that don't:

  • "Quick question" (overused, vague)
  • "Partnership opportunity" (screams template)
  • "Can I pick your brain?" (asks, doesn't give)

LinkedIn DM

Different channel, different rules. LinkedIn is conversational by design.

Before DMing:

  1. Engage with their content first (like, comment with substance)
  2. Wait a few days
  3. Then DM with reference to the engagement

LinkedIn DM structure:

  • Shorter than email (50-60 words)
  • More casual tone
  • Reference the content you engaged with

Example:

Hey [Name] — loved your post about [topic]. The point about [specific thing] is something I don't see people talk about enough.

We work with [similar companies] on exactly that problem. Would you be open to swapping notes?

Don't: Send LinkedIn DMs that read like formal emails. It's instant messaging, not correspondence.

Twitter/X DM

Twitter is the most casual. Build familiarity before DMing.

Warm-up sequence:

  1. Follow them
  2. Like/retweet a few things over a week
  3. Reply to a tweet with something substantive
  4. Then DM

Twitter DM style:

  • Very short (2-3 sentences)
  • Feels like a friend texting
  • Reference the tweet engagement

Example:

Hey — your thread on [topic] was great. We actually built something that solves the exact problem you mentioned in point 3. Worth sharing?


Anti-Patterns (Never Do These)

The Phrases That Kill

These instantly signal "template spam":

❌ "I hope this email finds you well"
❌ "I know you're busy, but..."
❌ "Just following up"
❌ "I wanted to reach out because..."
❌ "Let's circle back"
❌ "Would love to pick your brain"
❌ "I came across your profile and was impressed"
❌ "Are you the right person to talk to about..."

Why they fail: Everyone uses them. They broadcast "I sent this to 1,000 people."

The Structural Mistakes

Multiple CTAs:

"We could do a call, or I could send you a deck, or you could check out our site..."

Pick one. Multiple options = no action.

Leading with credentials:

"I'm the founder of X. We've raised $Y. We work with Z..."

Nobody cares who you are until they care what you can do for them.

The backhanded compliment:

"Your content is great, but I noticed your thumbnails could use work..."

Feels like negging. Compliment OR identify a problem, not both in the same sentence.

Too much too soon:

"I'd love to schedule a 45-minute strategy session to deep-dive into your growth challenges..."

You're asking for marriage on the first message. Ask for a coffee first.

The Tone Mistakes

Entitled:

"I'm sure you'll want to hear about this..."

Guilt-trippy:

"I've sent a few emails and haven't heard back..."

Fake urgency:

"I have a few spots left this month..."

Over-familiar:

"Hey buddy! Saw your stuff and had to reach out..."


Before/After Examples

Example 1: Agency Selling Video Production

Before (my default, spam-tier):

Subject: Quick question about your YouTube content

Hi [Name],

I came across your YouTube channel while researching companies in the [industry] space, and I was impressed by the valuable insights you're sharing.

I noticed your posting schedule has been inconsistent lately, and I think there might be an opportunity to significantly increase your channel's impact.

At [Agency], we specialize in helping SaaS founders create high-quality video content. We've helped companies like [Client] increase their subscribers by 300%.

I'd love to schedule a quick 15-minute call to discuss how we might help.

Best regards, [Name]

After (applying principles):

Subject: Your YouTube vs [Competitor]'s

[Name] — noticed you've posted 4 videos in the last 6 months while [Competitor] is doing 4/month. Your content is actually better, but they're winning on volume.

We helped [Similar Company] go from sporadic posting to weekly without adding headcount — just systematized their production.

Worth seeing how they did it?

What changed:

  • Specific observation (4 vs 4/month)
  • Competitive framing (creates urgency without fake urgency)
  • Give first (the insight itself is valuable)
  • Soft CTA (worth seeing vs. schedule a call)
  • 65 words vs 150

Example 2: Freelance Designer

Before:

Hi [Name],

I'm a freelance brand designer with 8 years of experience working with startups and scale-ups. I specialize in creating cohesive brand identities that help companies stand out.

I'd love to help [Company] elevate its visual presence. Would you be open to a quick call to discuss?

After:

[Name] — the product UI on [Company] is clean, but the marketing site feels like a different brand. Different typography, different color treatment, different energy.

I've helped 3 Series A companies close that gap in the last year. [Company X] is probably the closest comp to what you're building.

Want me to send what we did for them?

What changed:

  • Specific problem (UI vs marketing site mismatch)
  • Shows expertise through observation, not claims
  • Relevant social proof (Series A, similar company)
  • Give first (offering to share the case)

The Writing Process

  1. Research first (5-10 min per prospect for high-value targets)

    • Recent content/posts
    • Company news
    • Hiring signals
    • Obvious gaps or problems
  2. Find the hook — What specific thing makes outreach relevant RIGHT NOW?

  3. Write the give — What insight or value can you lead with?

  4. Draft short — Get it under 80 words

  5. Check for spam triggers — Any clichés? Multiple CTAs? Me-focused language?

  6. Soften the CTA — Make responding feel easy


Quality Checklist

Before sending:

  • Under 80 words?
  • Specific detail about THEM in first line?
  • Giving value before asking?
  • One CTA only?
  • CTA is soft (interest, not commitment)?
  • No spam-trigger phrases?
  • Would YOU reply to this?

That last one matters most. Read it as the recipient. Does it feel worth responding to, or does it feel like work?


What This Skill Changes

Without Skill With Skill
Generic opener Specific, researched hook
Me-focused pitch Them-focused relevance
Claims credentials Shows through specifics
Hard CTA (call request) Soft CTA (interest check)
150+ words Under 80 words
Template energy Human energy
1-3% reply rate 15-25%+ reply rate

The difference isn't tricks. It's doing the work to be relevant, then writing like a human.

Research Example: Building ai-creative-strategist

This document shows what good skill research looks like, using the ai-creative-strategist skill as a case study.


The Problem We Were Solving

Initial gap identified: Claude could generate images, but the output was generic. "Make me a social graphic" produced forgettable AI-slop. Even with good prompts, the creative THINKING was missing.

The expertise gap: A creative director doesn't just execute briefs. They:

  • Challenge the brief itself
  • Research before ideating
  • Explore multiple directions
  • Know WHY certain things work
  • Build systems, not one-offs

We needed to transfer this thinking, not just image generation techniques.


Research Phase 1: Visual Psychology

Questions we asked:

  • What makes people stop scrolling? (Attention psychology)
  • How does the brain process visual information?
  • What creates emotional response to images?

What we found:

Processing Order (Neurological)

  1. Faces (highest priority - neurologically wired)
  2. Movement/implied motion
  3. High contrast elements
  4. Unexpected/novel elements (pattern interrupts)
  5. Text (last - requires conscious processing)

The 3-Second Threshold

  • 65% of viewers who watch first 3 seconds continue to 10 seconds
  • First impression forms in 50 milliseconds (before conscious thought)
  • Implication: Visual must communicate BEFORE conscious processing

Pattern Interrupts That Work

  • Unexpected color combinations (intentional, not random)
  • Scale violations (tiny person, giant object)
  • Temporal mismatch (retro + futuristic)
  • Context displacement (product in impossible environment)

Sources: Academic psychology papers, UX research, advertising effectiveness studies


Research Phase 2: Direct Response Visual Principles

Questions we asked:

  • What did Ogilvy, Hopkins, Schwartz say about images?
  • What visual principles from direct response still work?
  • How do images and copy work together?

What we found:

Ogilvy's Rules (Still Work)

  1. Finished outcomes beat processes
  2. Specificity beats generality
  3. Reader-first composition (images → headlines → body)
  4. Authenticity over polish

Visual Hierarchy for Conversion

ATTENTION SEQUENCE:
1. PRIMARY FOCAL POINT (the hook)
   ↓
2. SUPPORTING CONTEXT (why care)
   ↓
3. CALL TO ACTION (what to do)

Hick's Law Applied

More choices = slower decisions = lower conversion

  • One dominant visual element
  • One clear message
  • One obvious next step

Sources: "Confessions of an Advertising Man" (Ogilvy), "Breakthrough Advertising" (Schwartz), conversion rate optimization studies


Research Phase 3: Anti-Generic AI Aesthetics

Questions we asked:

  • What makes AI images look "AI-ish"?
  • What patterns do experts flag as generic?
  • How do you avoid the statistical average?

What we found:

The Generic AI Markers

Visual tells:

  • Perfectly centered composition
  • Uniform, directionless lighting
  • Overly smooth/plastic textures
  • Symmetrical everything
  • "Too perfect" Disney aesthetic
  • Hands that don't look right

Aesthetic tells:

  • High saturation + no color strategy
  • Gradient backgrounds to nowhere
  • Generic "tech blue" palette
  • Stock photo poses

Why It Happens

AI models predict "most likely" output. Without specific direction, you get training data averages.

Generic prompts → generic results.

The Fix: Multi-Dimensional Style Architecture

Instead of single keywords, specify across dimensions:

STYLE ARCHITECTURE:
├── Lighting (direction, quality, temperature)
├── Color Palette (specific palette, not "colorful")
├── Composition (rule, framing, depth)
├── Artistic Reference (movement, artist, era)
├── Technical Treatment (camera, film stock, processing)
├── Texture/Surface (material quality, imperfections)
└── Emotional Register (specific mood, energy)

Sources: AI art community discussions, professional photographer critiques of AI output, brand creative director interviews


Research Phase 4: Leading Brand Visual Identities

Questions we asked:

  • What are the best tech/AI brands doing visually?
  • What makes each distinctive?
  • What can we learn from their approaches?

What we found:

Perplexity AI

  • Warm earth tones + techy dark blue
  • 3D metallic asterisk (sculptural, tactile)
  • Texture contrast (digital meets organic)
  • Why it works: Unexpected for AI. Creates tension that captures attention.

Anthropic Claude

  • Rust orange (#C15F3C) - notably NOT tech blue
  • Extreme restraint and simplicity
  • Rounded, humanistic typography
  • Why it works: Differentiates from every other AI company. Warmth = trust.

Figma

  • Dynamic primitives (jumbo shapes)
  • Motion as core expression
  • Expanded color palette
  • Why it works: Shows the product's creative potential through the brand itself.

Linear

  • Single desaturated blue
  • Extreme minimalism
  • Dark mode native
  • Why it works: Restraint = sophistication for developer audience.

Key insight: Each brand's visual identity communicates their positioning. The visuals ARE the message.

Sources: Brand guidelines, design analysis articles, case studies


Research Phase 5: Platform-Specific Optimization

Questions we asked:

  • What works on each platform?
  • What are the technical requirements?
  • What patterns drive engagement?

What we found:

Instagram

  • Carousels: 0.55% engagement (highest format)
  • 4:5 ratio for feed dominance
  • Consistent color palette across carousel
  • Swipe-worthy hooks on first slide

YouTube Thumbnails

  • Face + emotion + bright color + high contrast
  • 3-5 words max text
  • Curiosity gap (show result, hide method)
  • Background blur to isolate subject

LinkedIn

  • Professional but not corporate
  • Data visualization performs well
  • Dark blues = trust, professionalism
  • Multi-image carousels: 6.6% engagement

TikTok

  • Pattern interrupts in first frame
  • Authenticity over polish
  • Movement required
  • Text overlays for silent viewing

Sources: Platform data reports, creator case studies, algorithm analysis


The Discovery Phase: Style Exploration Process

This came from USING the skill, not just researching.

What happened: We kept generating images with similar styles (chrome, navy, Perplexity-like). The angles varied, but the STYLE was the same.

The breakthrough: When we forced genuinely different style territories, we found a winner (retro 90s nostalgia) that resonated completely differently.

What we learned:

Don't iterate style, explore it

WRONG: Generate 5 variations of "premium tech aesthetic"
RIGHT: Generate 5 genuinely DIFFERENT style territories

Style territories to explore:

Territory Description
Brutalist Stark, raw, anti-design
Retro Nostalgia 90s interfaces, vaporwave
Editorial Magazine aesthetic, sophisticated
Gradient Abstract Pure color fields + type
3D Premium Chrome, metallic, atmospheric
Illustrated Flat vector, playful
Photographic Real textures, lifestyle

When a style wins, extract principles

Don't repeat specific elements. Extract WHY it works.

Example - Retro Nostalgia principles:

PRINCIPLES (transferable):
├── Nostalgic tech interfaces (any era, any device)
├── Dialog/prompt/status format
├── Vaporwave palette (teal, purple, pink)
├── Ironic/playful tone
└── Retro UI elements

Then apply to different formats:

  • Desktop computer → Nokia phone → Game Boy → VHS tape → Arcade → iPod
  • Same vibe, endless variations = content SYSTEM

How Research Became Skill Structure

Research findings → Skill sections

Research Skill Section
Visual psychology "What Makes People Stop" (in VISUAL_INTELLIGENCE.md)
Direct response visual principles "Direct Response Visual Principles"
Anti-generic patterns "The Anti-Generic Imperative"
Brand case studies "Brand Aesthetic Patterns"
Platform optimization "Platform-Specific Considerations"
Style exploration process "The Visual Exploration Process" (added after testing)

What we cut (research that didn't make it)

  • Detailed color theory (too comprehensive, not actionable)
  • Art history deep dives (interesting but not useful for output)
  • Technical camera specifications (over-engineered for the use case)
  • Comprehensive platform algorithm breakdowns (became friction)

The test: Does this improve the output, or just make the skill more "complete"?


The Testing Phase

Test 1: Generate social graphics for thevibemarketer.com/skills

Without the style exploration section:

  • Generated chrome/navy Perplexity-style variations
  • Angles varied, style stayed the same
  • Output was fine but not distinctive

After adding style exploration:

  • Generated truly different territories
  • Found retro 90s aesthetic
  • Created a content SYSTEM (7+ format variations from one style)

Test 2: Prompt writing

Without natural language guidance:

"Premium, chrome, 4k, ultra detailed, professional, modern aesthetic"

Result: Generic AI output

After adding natural language section:

"A 1990s software box for 'Vibe Skills Pack' floating in dark space,
dramatic rim lighting catches the holographic disc surface creating
rainbow reflections, nostalgic 90s tech commercial aesthetic"

Result: Distinctive, ownable, systematic


Total Research Investment

Phase Time
Visual psychology ~4 hours
Direct response principles ~4 hours
Anti-generic patterns ~3 hours
Brand case studies ~4 hours
Platform optimization ~3 hours
Testing and iteration ~6 hours
Total ~24 hours

This doesn't include the original direct-response-copy research that informed the persuasion principles.


Key Takeaways for Your Research

  1. Research across multiple domains. ai-creative-strategist combined psychology, advertising, design, and platform optimization.

  2. Find primary sources. Ogilvy's actual book > blog post about Ogilvy.

  3. Identify anti-patterns. What makes output BAD is often more valuable than what makes it good.

  4. Test during research. The style exploration process wasn't in the original research - we discovered it by using the skill.

  5. Cut ruthlessly. We researched color theory extensively. None of it made it into the skill. It didn't improve output.

  6. Synthesize into principles. "Faces get attention" is a principle. "Use faces at 30-degree angles with catch lights in the eyes at f/2.8" is over-engineered noise.

The goal isn't comprehensive coverage. It's expertise transfer that improves output.

Research Methodology for Skill Creation

The depth of your research determines the quality of your skill. This document covers how to conduct research that actually makes skills better.


The Research Stack

You have access to different research tools depending on your environment:

Tier 1: Perplexity MCP (Best)

If you have the Perplexity MCP configured, use it. It's purpose-built for research.

Available tools:

  • perplexity_ask — Quick questions, conversational
  • perplexity_research — Deep research with citations
  • perplexity_reason — Reasoning through complex topics

When to use which:

Tool Use For
perplexity_ask Quick facts, definitions, "what is X"
perplexity_research Deep dives, "best practices for X", "how experts do X"
perplexity_reason Complex analysis, "why does X work", trade-off evaluation

Example research prompts for skill building:

perplexity_research: "What are the foundational frameworks and classic texts
for [domain]? I need primary sources, not summaries. Who are the foundational
thinkers and what are their key contributions?"

perplexity_research: "What separates expert-level [domain] work from amateur
work? What are the specific tells and patterns that indicate expertise?"

perplexity_research: "What are the most common mistakes in [domain]? What do
experts complain about when reviewing amateur work?"

perplexity_research: "What are the current best practices in [domain] as of 2024?
What has changed recently? What techniques are working now?"

Tier 2: Claude Web Search (Good Fallback)

If you don't have Perplexity, Claude's built-in web search works well.

How to use:

Simply ask questions that require current information. Claude will automatically use web search when needed.

"Search for the foundational frameworks and classic texts for direct response
copywriting. I need primary sources - books, essays, documented methodologies
from experts like Eugene Schwartz, Claude Hopkins, David Ogilvy."

"Search for what makes AI-generated content look generic. What are the specific
patterns that experts identify as 'AI slop'?"

"Search for the current best practices in YouTube thumbnail design. What's
working in 2024? What patterns do the highest-performing thumbnails share?"

Tips for better web search results:

  • Be specific about what you're looking for
  • Ask for primary sources, not summaries
  • Request recent information when relevance matters
  • Follow up on interesting sources mentioned in results

Tier 3: Manual Research (Always Valid)

Sometimes the best research is reading the actual books, not searching for summaries.

For copywriting, read:

  • "Breakthrough Advertising" by Eugene Schwartz
  • "Scientific Advertising" by Claude Hopkins
  • "The Boron Letters" by Gary Halbert
  • "The Adweek Copywriting Handbook" by Joseph Sugarman

For visual design, study:

  • Actual campaigns that performed (not "best of" lists)
  • Brand guidelines from companies you admire
  • Platform-specific content that gets engagement

The rule: Primary sources > summaries > blog posts > AI-generated summaries


The Research Process

Step 1: Define What You Need to Know

Before researching, list specific questions:

For direct-response-copy, we asked:

  • What are the classic frameworks that have stood the test of time?
  • What headline patterns consistently outperform others?
  • What makes copy feel "AI-generated" vs human-written?
  • What do the legends (Hopkins, Ogilvy, Halbert) agree on?
  • What modern internet-native patterns work now?

For ai-creative-strategist, we asked:

  • What makes people stop scrolling? (Visual psychology)
  • What patterns make AI images look generic? (Anti-patterns)
  • What visual styles are working commercially right now? (Current trends)
  • What do leading brands do with their visual identity? (Case studies)
  • What works on each platform? (Platform-specific)

Step 2: Find Foundational Sources

Every domain has foundational thinkers. Find them first.

Research prompt:

"Who are the foundational experts in [domain]? What are their seminal works?
I want the primary sources that practitioners still reference, not recent
summaries or interpretations."

What you're looking for:

  • Books/essays that defined the field
  • Frameworks that multiple experts reference
  • Principles that have stood the test of time
  • The "why" behind techniques, not just the "what"

Step 3: Find Real-World Examples

Theory without examples is useless. Find actual outputs.

For copy:

  • Landing pages that convert (with data if possible)
  • Ads that ran for decades (indicates performance)
  • Email sequences from successful creators
  • Sales pages that you've personally bought from

For visual:

  • Campaigns that won awards AND performed
  • Brand visual systems that are distinctive
  • Content that gets genuine engagement (not just likes)
  • Before/after brand evolutions

Research prompt:

"What are examples of excellent [domain] work? I need specific examples I can
study, not general descriptions. Bonus if there's data on performance."

Step 4: Identify Anti-Patterns

What makes output in this domain bad? This is often more valuable than knowing what's good.

Research prompt:

"What are the common mistakes and failure patterns in [domain]? What do experts
complain about when they see amateur work? What are the tells that someone
doesn't know what they're doing?"

For AI-generated content specifically:

"What patterns make AI-generated [content type] look generic or 'AI-ish'?
What are the specific tells that experienced people can spot?"

Step 5: Find Current Best Practices

Domains evolve. What worked in 2020 might not work now.

Research prompt:

"What are the current best practices in [domain] as of 2024? What has changed
recently? What new techniques are working? What old techniques have stopped
working?"

Step 6: Synthesize Into Principles

Don't just collect information. Synthesize it into principles.

The process:

  1. What do multiple sources agree on? (Likely fundamental)
  2. What do experts do that amateurs don't? (Expertise markers)
  3. What's the underlying "why"? (Transferable principle)
  4. What should be avoided? (Anti-patterns)

Research Quality Checklist

Before moving to skill writing, verify:

Breadth

  • Found foundational sources (books, frameworks, classic texts)
  • Found current best practices (recent, working now)
  • Found real examples (actual outputs, not descriptions)
  • Found anti-patterns (what to avoid)

Depth

  • Understand WHY techniques work, not just WHAT they are
  • Can explain the underlying principles
  • Know what separates good from great
  • Know common failure modes

Synthesis

  • Identified patterns across multiple sources
  • Distinguished fundamental vs trendy
  • Can articulate the "taste" of the domain
  • Have specific examples for key principles

Research Prompts by Skill Type

For Copy/Writing Skills

"What are the classic copywriting frameworks that have stood the test of time?
(AIDA, PAS, etc.) Who developed them and why do they work?"

"What headline patterns consistently outperform in direct response?
Cite specific examples and performance data if available."

"What makes modern internet-native copy different from traditional advertising?
What voice markers signal 'creator' vs 'corporate'?"

"What are the specific tells that copy was AI-generated?
What words, phrases, and patterns do experts flag?"

For Visual/Creative Skills

"What visual patterns stop scrolling? What does psychology research say about
attention and visual processing?"

"What makes AI-generated images look generic? What are the specific aesthetic
patterns that signal 'AI slop'?"

"What visual styles are working commercially in 2024? Cite specific brands
and campaigns."

"What are the platform-specific visual requirements for [Instagram/YouTube/LinkedIn]?
What performs best on each?"

For SEO/Content Skills

"What content structure patterns correlate with search ranking performance?
Cite studies or large-scale analyses."

"What are the current Google ranking factors that actually matter?
Distinguish confirmed from speculated."

"What makes content 'helpful' vs 'generic' according to Google's quality
guidelines? How do human raters evaluate?"

"What content formats are performing best for [topic/industry] in 2024?"

For Technical/Development Skills

"What are the established best practices for [technology/framework]?
Cite official documentation and expert consensus."

"What are common mistakes when implementing [feature/pattern]?
What do code reviewers frequently flag?"

"What architectural patterns are recommended for [use case]?
What are the trade-offs of each approach?"

Common Research Mistakes

Mistake 1: Stopping at Summaries

Wrong: Reading "10 copywriting tips" blog posts

Right: Reading Claude Hopkins' "Scientific Advertising" directly

Summaries lose nuance. Primary sources give you the thinking behind the technique.

Mistake 2: Collecting Without Synthesizing

Wrong: 50 pages of notes with no synthesis

Right: Clear principles extracted from research

Information isn't insight. You need to process research into transferable principles.

Mistake 3: Skipping Anti-Patterns

Wrong: Only researching "what works"

Right: Also researching "what fails"

Knowing what to avoid is often more valuable than knowing what to do. Claude's defaults often land in the "what fails" category.

Mistake 4: Ignoring Current Context

Wrong: Only reading classic sources

Right: Classic sources + current best practices

The fundamentals haven't changed, but execution has. A skill needs both.

Mistake 5: Research Without Testing

Wrong: Extensive research → immediately writing skill

Right: Research → test ideas → validate → write skill

Research tells you what might work. Testing tells you what actually works.


Research Time Investment

Skill Complexity Research Time
Simple (single technique) 1-2 hours
Medium (multiple techniques) 4-8 hours
Complex (full domain) 2-5 days

direct-response-copy took ~40 hours of research across classic texts, modern examples, and AI-specific anti-patterns.

ai-creative-strategist took ~20 hours of research on visual psychology, brand case studies, and platform optimization.

This investment pays off. Shallow research = shallow skill = mediocre output.


The Meta-Principle

Research isn't about collecting information. It's about developing taste.

When you've done enough research, you should be able to:

  1. Identify good work instantly
  2. Explain WHY it's good
  3. Spot bad work and name the problems
  4. Know what Claude would get wrong without guidance

That's when you're ready to write the skill.

Vibe Skill Creator

Build Claude skills that actually improve output. Includes a finished example you can use immediately.


Start Here

New to skills? Read START-HERE.md first (3-minute read).

It explains:

  • What skills are and why they matter
  • How to install them in Claude Desktop or Claude Code
  • How to get started immediately

What's Included

File Purpose Use It For
START-HERE.md Quick intro + installation Read this first
SKILL.md The skill-building skill Creating your own skills
cold-outreach/SKILL.md Finished example skill Writing cold emails that get replies
references/ Deep-dive materials Going deeper (optional)

Two ways to use this:

  1. Build skills — Use SKILL.md to create skills for your domain
  2. Use the example — Drop cold-outreach/SKILL.md into Claude and write better outreach immediately

Quick Start

To build a skill:

Upload SKILL.md to Claude and say:

I want to build a skill for [YOUR DOMAIN]. Guide me through the process —
explore where you fail without guidance, research the domain, then help me
build a skill that actually improves your output.

To write cold outreach:

Upload cold-outreach/SKILL.md to Claude and say:

Write a cold email to [PROSPECT] about [YOUR SERVICE].

The 10-Step Process

When building a skill, Claude guides you through:

1. UNDERSTAND → What skill? What problem?
2. EXPLORE    → See where Claude fails without guidance
3. RESEARCH   → Go deep on the domain
4. SYNTHESIZE → Extract principles from research
5. DRAFT      → Write initial skill
6. CRITIQUE   → Review against quality criteria
7. ITERATE    → Fix gaps, get feedback
8. TEST       → Use on a real scenario
9. FINALIZE   → Structure properly
10. REFERENCE → Build deep expertise docs (if needed)

You don't need to know how skills work. Just have the conversation.


Study the Example

cold-outreach/SKILL.md demonstrates:

  • Expert voice (not documentation)
  • Principles with WHY (not just WHAT)
  • Anti-patterns explicitly named
  • Real before/after examples
  • Quality checklist

It's also immediately useful — drop it into Claude and write better cold emails.


Common Mistakes to Avoid

The references/bad-skill-patterns.md file documents 10 patterns that make skills mediocre:

  1. Documentation voice (sounds like a wiki)
  2. Procedure over principle (steps without thinking)
  3. Comprehensive coverage (information overload)
  4. Generic tool references (vague instructions)
  5. Missing anti-patterns (only what TO do)
  6. Friction creation (intermediate documents)
  7. Over-engineered structure (complexity without value)
  8. Theoretical examples (fake scenarios)
  9. Missing the WHY (rules without reasons)
  10. "Just in case" additions (bloat)

Read it before building your first skill.


Go Deeper

This package includes one skill-building skill + one example. Two ways to get more:


Option 1: Get the Full Skills Pack

The complete Vibe Skills Pack includes 15+ production-ready skills:

Skill What It Does
Direct Response Copy Write landing pages, emails, and ads that convert
Lead Magnet Creation Design lead magnets that build lists and bridge to paid
Email Sequences Build welcome, nurture, and conversion sequences
SEO Content Create content that ranks AND reads like a human wrote it
Newsletter Writing Write newsletters people actually want to read
Positioning & Angles Find the angle that makes something sell
Content Atomization Turn one piece into platform-optimized assets
Keyword Research Strategic keyword research without expensive tools
AI Image Generation Generate images with optimal quality
AI Product Photography Professional product shots and hero images
AI Social Graphics Platform-optimized graphics and thumbnails
AI Product Video Product reveals and animated shots
AI Talking Head Presenter videos and lip-synced avatars
Brand Voice Define or extract consistent brand voice
Orchestrator Routes you to the right skill for any marketing task

Each skill is battle-tested with deep reference material.

Get the skills pack: thevibemarketer.com/skills


Option 2: Join the Community (Best Value)

The Vibe Marketer community is where these skills come from.

What you get:

  • 50% off the skills pack (members-only discount)
  • Access to 2,800+ marketers building with AI
  • Weekly skill updates and improvements
  • Direct access to skill creators
  • Share what you build, get feedback
  • See how others are using skills

$199/year — Join 2,800+ members

thevibemarketer.com


Questions? Feedback? Join the community or reach out on Twitter @thevibemarketer

Start Here: What Are Claude Skills?

Read time: 3 minutes


The Problem

You ask Claude to write a cold email. It produces this:

"I hope this email finds you well! I came across your company and was impressed by your innovative approach. I'd love to schedule a quick 15-minute call to discuss how we might help you achieve your goals..."

Generic. Template-sounding. The kind of email everyone deletes.


The Solution: Skills

A skill is a document that gives Claude domain expertise. Not instructions — expertise. The thinking patterns, principles, and anti-patterns that separate amateurs from professionals.

With the cold-outreach skill loaded, Claude produces this instead:

"[Name] — noticed you've posted 4 videos in the last 6 months while [Competitor] is doing 4/month. Your content is actually better, but they're winning on volume.

We helped [Similar Company] go from sporadic posting to weekly without adding headcount.

Worth seeing how they did it?"

Same AI. Different expertise.


What's In This Package

File What It Does
START-HERE.md You're reading it
SKILL.md The skill that teaches you to build skills
cold-outreach/SKILL.md A finished example skill (use it or study it)
references/ Deep-dive materials for skill builders

How to Use This

Option 1: Build Your Own Skills

Use SKILL.md to create skills for your domain. The skill guides you through a 10-step process — you don't need to know how skills work. Just have the conversation.

Start by saying:

"Help me build a skill for [your domain]"

Option 2: Use the Cold Outreach Skill

Drop cold-outreach/SKILL.md into Claude and immediately write better outreach.

Option 3: Both

Build your own skills AND use the cold-outreach skill. They work independently.


Installation

For claude.ai (Claude Desktop)

Method A: Upload as File (Easiest)

  1. Start a new chat
  2. Click the attachment icon (📎)
  3. Upload SKILL.md (or cold-outreach/SKILL.md)
  4. Start chatting — the skill is now active for this conversation

Method B: Add to a Project (Persistent)

  1. Create a new Project (or open existing)
  2. Click "Project knowledge" → "Add content"
  3. Paste the contents of the skill file
  4. The skill stays active for all chats in that project

Method C: Install as Custom Skill (Permanent)

  1. Create a folder named vibe-skill-creator (or cold-outreach)
  2. Put SKILL.md inside the folder
  3. ZIP the folder
  4. Go to Settings → Features → "Custom skills"
  5. Click "Upload skill" and select your ZIP
  6. The skill appears in your Skills list and activates automatically when relevant

For Claude Code (Terminal)

Method A: Project Skill (For Teams)

# From your project root:
mkdir -p .claude/skills/vibe-skill-creator
cp SKILL.md .claude/skills/vibe-skill-creator/

# Commit to git — teammates get it automatically

Method B: Personal Skill (Just You)

mkdir -p ~/.claude/skills/vibe-skill-creator
cp SKILL.md ~/.claude/skills/vibe-skill-creator/

# Available in all your Claude Code projects

Method C: Slash Command (Manual Trigger)

mkdir -p .claude/commands
cp SKILL.md .claude/commands/build-skill.md

# Invoke with: /build-skill

How Claude Code Skills Work:

  • Skills are auto-invoked based on the description in the frontmatter
  • You don't call them explicitly — just work normally
  • Claude detects when the skill is relevant and uses it

Quick Start

To build a skill:

I want to build a skill for [YOUR DOMAIN]. Guide me through the process —
explore where you fail without guidance, research the domain, then help me
build a skill that actually improves your output.

To use cold outreach:

Write a cold email to [PROSPECT] about [YOUR SERVICE].
Context: [Any relevant details about the prospect]

What Makes This Different

Most "prompt engineering" teaches you tricks. This teaches you how to transfer expertise.

The difference:

  • Tricks: "Use these magic words for better output"
  • Skills: "Give Claude the thinking patterns experts use"

Skills compound. One well-built skill improves hundreds of future outputs.


Next Steps

  1. Try the cold-outreach skill — see the difference immediately
  2. Build one skill for something you do repeatedly
  3. Study the references if you want to go deeper

Want More?

This is one skill-building skill + one example. Here's how to go deeper:

Option 1: Get the Full Skills Pack

15+ production-ready skills: Direct Response Copy, Lead Magnets, Email Sequences, SEO Content, Newsletters, AI Image Generation, AI Product Photography, and more.

Each skill is battle-tested with deep reference material.

Get the skills pack: thevibemarketer.com/skills


Option 2: Join the Community (Best Value)

The Vibe Marketer community is where these skills come from.

What you get:

  • 50% off the skills pack (members-only discount)
  • Access to 2,800+ marketers building with AI
  • Weekly skill updates and improvements
  • Share what you build, get feedback
  • Direct access to skill creators

$199/yearthevibemarketer.com


Built by The Vibe Marketer community — 2,800+ marketers building with AI.

name description
vibe-skill-creator
Build world-class skills through guided conversation and recursive self-review. Use when creating or improving a skill. Triggers on: build skill, create skill, improve skill, help me make a skill. Guides users through discovery, research, drafting, testing, and iteration until the skill actually works.

Vibe Skill Creator

Most skills are mediocre because people don't know what good looks like, what questions to ask, or how to improve what they have.

This skill fixes that. It guides you through building a skill from scratch — exploring the problem, researching the domain, drafting, self-critiquing, testing with real scenarios, and iterating until it actually works.

You don't need to know how to build a skill. You need to have a conversation, and the skill emerges.


Core Truths (Non-Negotiable)

Before building anything, internalize these:

Truth 1: Expertise Transfer, Not Instructions

A skill should make Claude think like an expert, not follow rules.

Documentation (wrong):

"Step 1: Write a headline. Step 2: Write body copy. Step 3: Add CTA."

Expertise (right):

"The headline does 80% of the work. One headline can outpull another by 19.5x. Here's what separates the winners..."

Truth 2: Flow, Not Friction

Flow skills produce output. User asks → skill delivers → user can act immediately.

Friction skills create intermediate documents. Research reports, comprehensive analyses, strategic frameworks that feel obligatory to incorporate.

The Audience-Intel Lesson: We built a comprehensive audience research skill. Excellent output — segments, pain points, language patterns. Then we used it to write copy. The copy was worse than copy written without the research.

Why? Comprehensive research creates pressure to use it all. Cover all segments. Address all pain points. Be thorough. Result: verbose, diluted, trying to do everything.

The lesson: More information ≠ better output. Skills should enable, not obligate.

Truth 3: Voice Matches Domain

A copywriting skill sounds like a copywriter. A creative skill sounds like a creative director. A cold outreach skill sounds like someone who actually sends cold emails.

If the skill reads like documentation, it's wrong.

Truth 4: Focused Beats Comprehensive

A skill for "cold outreach for selling services" beats a skill for "all sales communication." Constrain ruthlessly. Every section must earn its place.


Works for Any Domain

The process below works for any knowledge work — not just marketing.

Example domains and what exploration reveals:

Domain What Claude Gets Wrong (Default) What the Skill Would Fix
Code Documentation Describes WHAT the code does, not WHY it exists or when to use it. Documentation reads like a reference manual nobody would actually read. Write docs like a senior engineer explaining to a new teammate. Lead with the problem it solves, not the API surface.
Investor Updates Buries the lead with background context. Reads like a report, not a conversation. No clear ask or next step. Open with the one thing they need to know. Bad news first, framed with what you're doing about it. Every update ends with a specific ask.
Technical RFCs Comprehensive to the point of unreadable. Lists every consideration, no opinion on tradeoffs. Missing the "why not" for rejected alternatives. Strong recommendation upfront. Rejected alternatives with honest tradeoffs. Written for the skeptic, not the supporter.
Legal Briefs Generic structure, buries the strongest argument. Doesn't anticipate counterarguments. Reads like a template. Lead with your single best argument. Acknowledge and dismantle the other side's strongest point early. Every sentence earns its place.
Product Specs Feature list without user motivation. No success criteria. Doesn't constrain scope. Start with the user problem. Define what "done" looks like. Explicitly list what's NOT in scope.
Academic Writing Passive voice, hedging everywhere, buries the contribution. Sounds like every other paper. State your contribution in the first paragraph. Active voice. One idea per sentence.

The process is identical: Explore what Claude gets wrong → Research what experts do → Extract principles → Draft → Critique → Test → Iterate.

The cold-outreach example below shows the depth. Apply the same depth to your domain.


The Process: Start to Finish

Here's the full flow for building a skill from scratch:

┌─────────────────────────────────────────────────────────┐
│                                                         │
│   1. UNDERSTAND → What skill? What problem?             │
│          ↓                                              │
│   2. EXPLORE → See where Claude fails without guidance  │
│          ↓                                              │
│   3. RESEARCH → Go deep on the domain                   │
│          ↓                                              │
│   4. SYNTHESIZE → Extract principles from research      │
│          ↓                                              │
│   5. DRAFT → Write initial skill                        │
│          ↓                                              │
│   6. SELF-CRITIQUE → Review against quality criteria    │
│          ↓                                              │
│   7. ITERATE → Fix gaps, get feedback, improve          │
│          ↓                                              │
│   8. TEST → Use skill on a real scenario                │
│          ↓                                              │
│   9. FINALIZE → Codify into optimal structure           │
│                                                         │
└─────────────────────────────────────────────────────────┘

Step 1: Understand

Start by getting clear on what you're building.

Questions to answer:

Question Why It Matters
What domain? "Email marketing" is too broad. "Cold outreach for selling services" is focused.
What problem does this solve? Not "what does it do" — what does Claude get WRONG without it?
What does good look like? Find examples of excellent work in this domain.
What's the output? Landing page? Email? Strategy? Be specific.

Ask the user:

  • "What are you trying to produce with this skill?"
  • "What have you tried? What wasn't good enough?"
  • "Do you have examples of excellent work in this domain?"

If you can't articulate the problem clearly, you're not ready to move forward.


Step 2: Explore (See Where Claude Fails)

Before researching or writing anything, try to produce the output together.

This is crucial. You need to see what Claude gets wrong without guidance.

How to explore:

"Let me try writing a [cold email / landing page / whatever] without any special guidance. I want to see my default output so we know what the skill needs to fix."

Then produce the output. It will probably be bad. That's the point.

Example from building cold-outreach skill:

We asked Claude to write a cold email for a video production agency. The default output:

"Hi [Name], I came across your YouTube channel while researching companies in the [industry] space, and I was impressed by the valuable insights you're sharing. I noticed your posting schedule has been inconsistent lately... At [Agency], we specialize in helping SaaS founders create high-quality video content. I'd love to schedule a quick 15-minute call to discuss..."

What we identified:

  • Generic opener ("I came across your...")
  • Backhanded compliment ("impressed... BUT inconsistent")
  • Me-focused pitch (talks about the agency, not the prospect)
  • Hard CTA ("schedule a 15-minute call")
  • Too long (150+ words)

This exploration told us exactly what the skill needed to fix. Without it, we'd be guessing.

Document what you find:

  • What patterns does Claude default to?
  • What feels generic or AI-generated?
  • What would an expert do differently?

Step 3: Research (Go Deep)

Now research the domain. This is where most skill-builders cut corners. Don't.

Best approach: Use a subagent with web search.

Launch a research subagent (in Claude Code) or do multiple web searches to find:

  1. Foundational experts — Who are the recognized authorities? What frameworks do they teach?
  2. What actually works — Case studies, real examples, specific patterns that succeed
  3. Anti-patterns — What makes output in this domain bad? What are the tells?
  4. The WHY — Why do certain techniques work? What's the psychology?

Research prompt for subagent:

Research [domain] deeply. I need to understand what separates expert-level
work from amateur work.

Find:
1. Who are the foundational experts/thinkers? What frameworks do they teach?
2. What actually works? Case studies, real examples, specific patterns.
3. What are the anti-patterns? Common mistakes, things that make output feel generic/AI.
4. Why do the good techniques work? What's the underlying psychology?

Search for blogs, case studies, expert content, community discussions.
Return a synthesis of principles, not just a list of facts.

Example from cold-outreach skill:

Research found:

  • Alex Berman's 3C's: Compliment → Case Study → CTA
  • Josh Braun's 4T: Their Job → What's Hard → Your Alternative → Ask
  • Will Allred's Mouse Trap: Personalized trigger + value prop question
  • Key data: Each long sentence reduces reply rate by 17%
  • Anti-patterns: "I hope this email finds you well," "Just following up," etc.

This research became the foundation of the skill.

No MCP? Use web search directly:

Search: "best cold email examples that got replies"
Search: "cold outreach experts frameworks"
Search: "what makes cold email feel like spam"
Search: "[domain] common mistakes"
Search: "[expert name] [domain] framework"

Go deep. Surface-level research = surface-level skill.


Step 4: Synthesize

Turn research into principles.

Don't just list what you found. Extract the WHY behind it.

From research to principle:

Research Finding Principle
"Emails under 80 words get more replies" Short beats long — People scan, not read. If they can't understand in 2 seconds, they filter.
"Alex Berman uses Compliment → Case Study → CTA" Give before ask — Reciprocity is hardwired. Lead with value, not request.
"Experts say 'I hope this email finds you well' kills response rate" Anti-pattern: template openers — They signal "I sent this to 1,000 people."

The synthesis becomes the core of your skill.


Step 5: Draft

Now write the initial skill. It won't be perfect. That's fine.

Focus on:

  • Capturing what worked in exploration
  • Writing in the domain's voice (not documentation voice)
  • Stating principles with WHY, not just WHAT
  • Including concrete examples (before/after)
  • Explicitly listing anti-patterns

Don't aim for comprehensive. Aim for a working first draft.

Structure for first draft:

# Skill Name

[Opening that establishes voice and frames the problem]

## Why [Default Approach] Fails

[What Claude gets wrong, based on exploration]

## Core Principles

[Principles from research, with WHY]

## Anti-Patterns

[What to avoid, explicitly]

## Examples

[Before/after showing the principles in action]

Step 6: Self-Critique

This is the key step. Review your draft honestly against these criteria:

Voice Check

  • Does this sound like an expert in the domain?
  • Or does it read like documentation / a wiki article?

If documentation voice, rewrite with personality.

Principles Check

  • Am I explaining WHY techniques work, not just WHAT to do?
  • Could Claude adapt these principles to new situations?

If it's procedures without reasoning, add the WHY.

Anti-Pattern Check

  • Have I explicitly named what to AVOID?
  • What does Claude get wrong by default?

If no anti-patterns, add them. Often the most valuable part.

Example Check

  • Do I have concrete before/after examples?
  • Do they demonstrate the principles viscerally?

If examples are theoretical, make them concrete.

Focus Check

  • Does every section earn its place?
  • Would removing any section make output worse?

If anything is "just in case," cut it.

Example self-critique:

Draft 1 Assessment:

Voice: Good — sounds like someone who sends cold outreach

Principles: Strong — explains WHY short emails work, WHY soft CTAs convert

⚠️ Anti-patterns: Present but could be more explicit

Examples: Missing follow-up sequence examples

Missing: Trigger events section (user mentioned "timely" as key)

Priority fixes: Add follow-up examples, add trigger events section


Step 7: Iterate

Fix the gaps you identified. Get user feedback. Improve.

The loop:

  1. Fix priority issues from self-critique
  2. Present to user with honest assessment: "Here's what I improved, here's what's still weak"
  3. Get feedback
  4. Incorporate feedback (if it improves output, not just adds complexity)
  5. Self-critique again
  6. Repeat until it passes review

Present with honesty:

"Here's draft 2.

Improved: Added complete follow-up sequence with 4 email examples. Added trigger events section with 6 high-intent signals.

Still weak: LinkedIn section is shorter than email section — might need more examples.

Questions:

  • Does this cover the scenarios you care about?
  • Any anti-patterns I'm missing?"

Step 8: Test (Critical)

Before finalizing, test the skill on a real scenario.

This is non-negotiable. You don't know if the skill works until you use it.

How to test:

"Give me a real scenario — a specific service you'd sell to a specific type of prospect. I'll write outreach using the skill so we can see if it actually works."

Then produce output WITH the skill loaded. Compare to what you produced in Step 2 (exploration).

Example from cold-outreach skill:

Scenario: Selling AI automation services to a small law firm doing personal injury.

Output with skill:

Sarah — saw you're hiring a paralegal. Guessing intake volume is part of why.

Quick observation: your contact form goes to a generic inbox. Morgan & Morgan's triggers an instant callback + text sequence + case qualifier — all automated.

We built something similar for [Small PI Firm] — went from 30% lead response rate to 85% without adding staff.

Worth a 10-min look?

What the skill changed:

  • Specific trigger (hiring → intake volume)
  • Competitive comparison (Morgan & Morgan)
  • Concrete proof (30% → 85%)
  • Soft CTA ("Worth a 10-min look?")
  • 71 words (not 150)

The test passed. Output was genuinely better than the default.

If the test fails: Go back to Step 6, identify what's missing, iterate.


Step 9: Finalize

When the skill passes testing, structure it properly.

Optimal Skill Structure

---
name: skill-name
description: "One-line description. Use when [trigger]. Produces [output]."
---

# Skill Name

[Opening that establishes voice and frames the problem — 2-3 paragraphs max]

---

## Why [Default Approach] Fails

[What Claude gets wrong without this skill]

---

## Core Principles

[The fundamental truths of this domain — with WHY]

---

## Frameworks (if applicable)

[Specific frameworks from experts, with examples]

---

## Anti-Patterns

[What to avoid — explicit list with examples]

---

## Examples

[Real before/after demonstrating the principles]

---

## Quality Checklist

[Quick checklist for reviewing output]

Length Guidelines

  • SKILL.md: Under 500 lines
  • Complex domains: Use references/ folder for deep material
  • Simple distribution: Can compile into single file for claude.ai

Step 10: Build Reference Material (If Needed)

Some skills are self-contained. Cold-outreach fits in ~480 lines because the domain is focused.

Other skills need deep reference material. Visual creative strategy, direct response copywriting, SEO content — these domains have too much expertise to fit in a single file.

When You Need References

Skill Type References Needed?
Focused domain (cold outreach, lead magnets) Usually no — self-contained
Deep expertise domain (visual design, copywriting) Yes — vocabulary, examples, psychology
Tool-heavy domain (image generation, video) Yes — model IDs, prompts, technical specs
Multi-format domain (content atomizer) Yes — platform-specific rules

What Good References Contain

Not just information — transferable expertise.

Reference material should include:

  1. Domain vocabulary — The specific terms experts use

    • Visual creative: "chrome aesthetic," "biomorphic," "pattern interrupt"
    • Copywriting: "curiosity gap," "specificity," "fascination"
  2. Psychology/principles — WHY techniques work

    • "Faces get processed before anything else — neurologically wired"
    • "Each long sentence reduces reply rate by 17%"
  3. Examples from the best — What world-class looks like

    • Brand case studies (Perplexity, Notion, Linear)
    • Before/after transformations
    • Real campaigns that worked
  4. Anti-patterns with specifics — What to avoid

    • "Generic AI markers: perfectly centered, uniform lighting, plastic skin"
    • "Phrases that kill: 'I hope this email finds you well'"
  5. Templates/frameworks — Reusable structures

    • Prompt templates for image generation
    • Email frameworks (3C's, 4T, Mouse Trap)

Example: What Reference Material Looks Like

Example 1: ai-creative-strategist (separate reference files)

The VISUAL_INTELLIGENCE.md reference (~730 lines) includes:

  • Visual psychology (attention hierarchy, 3-second threshold)
  • Direct response principles (Ogilvy rules that still work)
  • Anti-generic AI techniques (how to avoid "AI slop")
  • Style vocabulary taxonomy (chrome/metallic, organic, surrealist, etc.)
  • Leading brand breakdowns (Perplexity, Anthropic, Figma, Notion)
  • Platform-specific strategies (Instagram, YouTube, TikTok, LinkedIn)
  • Prompt templates for different use cases

Example 2: direct-response-copy (embedded reference material)

The skill has ~500 lines of core content, then ~1,600 lines of reference material embedded after a divider:

---
---

# REFERENCE MATERIAL

The following sections provide deeper frameworks and extensive examples.

Reference includes:

  • Classic frameworks from Schwartz, Hopkins, Ogilvy, Halbert, Caples, Sugarman, Collier
  • Headline formulas with extensive examples
  • Opening lines and hooks (15+ patterns)
  • Curiosity gaps and open loops
  • Flow techniques (the slippery slide)
  • Modern internet-native copy examples

When to use which approach:

Approach Best For
Separate files Multiple reference topics, easier to update independently, cleaner main skill
Embedded after divider Single cohesive reference, simpler distribution (one file), easier to share

Both work. The key is having the deep expertise somewhere — not whether it's in separate files or appended to the skill.

How to Build References

During research (Step 3), collect:

  • Expert frameworks and methodologies
  • Specific vocabulary they use
  • Examples of excellent work
  • Data and psychology behind techniques
  • Platform or tool-specific details

After drafting, ask:

  • What expertise is too deep for the main skill?
  • What vocabulary does Claude need to sound like an expert?
  • What examples would help Claude understand "what good looks like"?

Structure references by topic:

references/
├── visual-intelligence.md      # Deep expertise on visual design
├── platform-strategies.md      # Platform-specific rules
├── prompt-templates.md         # Reusable prompt structures
└── brand-examples.md           # Case studies of excellent work

Finding Reference Material

The best sources for building references:

Source Type What You Get How to Find
Expert blogs Frameworks, principles "[domain] best practices blog"
YouTube breakdowns Real examples, analysis "[expert name] [domain] tutorial"
Communities What actually works Reddit, Twitter, Discord for the domain
Case studies Proof of results "[brand] case study [domain]"
Books/courses Deep methodology Find the foundational texts

The goal: Collect enough expertise that Claude can think like a domain expert, not just follow rules.


The Full Process: Real Example

Here's how we built the cold-outreach skill:

Step What We Did
1. Understand User wanted cold outreach for selling services. Multi-channel (email, LinkedIn, Twitter).
2. Explore Claude wrote default cold email — generic, long, hard CTA. Identified exactly what was wrong.
3. Research Subagent searched for Alex Berman, Josh Braun, Will Allred frameworks. Found anti-patterns, psychology, data.
4. Synthesize Extracted principles: research = relevance, give before ask, short beats long, soft CTA, follow up.
5. Draft Wrote initial skill with principles, frameworks, anti-patterns, before/after examples.
6. Self-Critique Identified gaps: missing follow-up examples, missing trigger events section.
7. Iterate Added 4-email follow-up sequence, added trigger events table with example hooks.
8. Test Real scenario: AI automation → law firm. Output was good. Skill passed.
9. Finalize Structured into optimal format, ~480 lines.

Time: ~1 hour conversation

Result: Skill that transforms Claude's cold outreach from spam-tier to actually good.


What Claude's Default Gets Wrong

When building skills without this process, I produce:

My Default What Actually Works
Documentation voice Domain expert voice
Comprehensive coverage Ruthless focus
"Step 1, Step 2, Step 3" Principles that transfer thinking
Theoretical examples Real before/after
What to do Why it works + what to avoid
Generic tool references Specific tool constraints
Everything "just in case" Only what improves output
Skip testing Test with real scenario

This skill exists to override those defaults.


Skills vs Subagents

Not everything should be a skill.

Skills = Execution (works in claude.ai AND Claude Code)

  • Direct production of outputs
  • Taste and principles baked in
  • User asks → skill produces

Subagents = Research (Claude Code only)

  • Go deep on analysis
  • Return findings, then get out of the way
  • Great for the research phase of skill-building
Use Case Better As
"Write cold outreach" Skill
"Research cold outreach best practices" Subagent
"Create SEO content" Skill
"Analyze competitor positioning" Subagent

Self-Review Checklist

Before finalizing any skill:

Focus

  • Scope is clearly constrained
  • Every section earns its place
  • Under 500 lines (references separate)

Voice

  • Sounds like an expert, not documentation
  • Has personality and point of view

Principles

  • Teaches WHY, not just WHAT
  • Claude could adapt to new situations

Anti-Patterns

  • Explicitly names what to avoid
  • Prevents common AI tells

Examples

  • Real before/after transformations
  • Contrast is visceral

Tested

  • Used on a real scenario
  • Output was genuinely better

References

For deeper methodology:

  • references/research-methodology.md — How to research using subagents and web search
  • references/research-example-ai-creative.md — Real example: research that built ai-creative-strategist
  • references/bad-skill-patterns.md — What mediocre skills look like (avoid these)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment