AI Skill Writer

Description

Turn a messy prompt you keep reusing into a clean, structured skill that anyone can copy and get consistent results. You paste what you currently tell AI — this skill restructures it so the output is reliable every time, not just when you happen to phrase it right.

When to Use

• You've got a prompt you copy-paste into ChatGPT/Claude every week and you're tired of tweaking it each time
• Your team keeps asking "what's the prompt for [X]?" and you want to give them something better than a Slack message
• You want to publish a skill on LIST and need it to follow the standard format

What Makes a Skill Different from a Prompt

A prompt is a one-shot instruction: "Write me a product review." Sometimes it works great, sometimes it doesn't.

A skill is a repeatable system: it tells the AI what inputs to collect, what steps to follow, what format to output, and what pitfalls to avoid. The difference is consistency — a well-written skill produces 8/10 output every time instead of alternating between 10/10 and 4/10.

The key elements that make skills reliable:

1. Specific instructions — "Write a review" fails. "Lead with the reader's pain point, then introduce the product as the solution, structure with H2 headings targeting '[product] review [year]'" succeeds.
2. Output format with structure — Show the AI exactly what the output should look like. A code block template is 10x more effective than describing the format in prose.
3. A real example — One concrete input/output pair grounds the AI's understanding better than 500 words of explanation.
4. Tips that prevent common failures — If you know the AI tends to be too generic, write "Include specific numbers and timeframes, not vague claims like 'saves a lot of time'."

Instructions

1. Ask for either: the raw prompt they currently use, OR a description of what they repeatedly ask AI to do

2. Analyze what the prompt actually does:

   • What's the task? (content creation, research, analysis, planning?)
   • Who's it for? (what audience or role?)
   • What inputs vary each time? (product name, audience, URL, etc.)
   • What does good output look like vs bad output?

3. Restructure into the standard skill format using these 7 sections:

Description (2 sentences max) First sentence: what it does. Second sentence: who it's for or when to use it. This shows as the preview text in the feed — make it specific enough that someone knows if it's relevant to them.

Bad: "Helps with content creation." Good: "Write product comparison posts that rank for '[X] vs [Y]' keywords. Built for affiliate bloggers who want to capture bottom-of-funnel search traffic."

When to Use (3 bullets) Three specific scenarios, not generic ones. "Writing a blog post" is useless. "You need to publish a comparison post for two competing SaaS tools this week" is useful.

Instructions (numbered, specific) This is the core. Every step should be concrete enough that the AI can't misinterpret it. Include:

• What to do (action)
• How to do it (specific approach)
• What good looks like (quality bar)

Bad: "3. Write the pros and cons" Good: "3. Write at least 3 pros and 2 cons. Each pro/con should be one specific feature or limitation, not a vague category. 'Exports to 12 formats including PDF and DOCX' is good. 'Great export options' is bad."

Input Required (bold labels) List every variable input with a label and a one-line description. Mark optional inputs.

Output Format (code block) Show the exact structure in a markdown code block. Use placeholders in brackets: [Product Name], [Price], etc. This is the single most important section — it determines output consistency.

Example (concrete input/output) One real input with a real (or realistic) output excerpt. Don't just describe the output — show 3-5 lines of what it actually looks like.

Tips (3-5 practical) What goes wrong when someone uses this skill poorly? Write tips that prevent those specific failures.

4. Test the skill: Copy the full output, paste it into ChatGPT or Claude with a test input, and verify the output matches expectations. If it doesn't, refine the Instructions or Output Format.

Input Required

• Your current prompt or workflow — paste the exact text you use, or describe what you ask AI to do repeatedly
• What goes wrong (optional) — when the output is bad, what's the usual problem? (too generic, wrong format, misses key info, etc.)

Output Format

## Description

[1-2 sentences. Specific. Shows as preview text in feed.]

## When to Use

- [Specific scenario 1]
- [Specific scenario 2]
- [Specific scenario 3]

## Instructions

1. [Concrete step with action + approach + quality bar]
2. [Next step]
3. [Next step]
...

## Input Required

- **[Input 1]** — [what it is and why it's needed]
- **[Input 2]** — [what it is]
- **[Input 3]** (optional) — [what it is]

## Output Format

[Markdown code block showing exact structure with [placeholder] brackets]

## Example

**Input**: [Real, specific input]

**Output excerpt**:
> [3-5 lines of actual output showing what the skill produces]

## Tips

- [Tip that prevents a specific common failure]
- [Tip that improves output quality]
- [Tip about when NOT to use this skill]

Example

Input: "I always ask ChatGPT: write me a comparison blog post about Notion vs Obsidian for my affiliate site, include pros cons pricing and a recommendation with my affiliate links. Sometimes it's great, sometimes it writes a generic listicle."

Output excerpt:

Description

Write detailed product comparison posts that rank for "[X] vs [Y]" keywords and convert readers through balanced, honest analysis with affiliate CTAs. Built for affiliate bloggers targeting bottom-of-funnel search traffic.

Instructions

1. Research both products: pricing (exact plans and prices), key features (top 5 each), ideal user profile, and one genuine weakness per product
2. Open with the reader's decision context: "You're choosing between [X] and [Y] for [use case]. Here's what matters." Not "In today's digital world, choosing the right tool..."
3. Build a comparison table with 6-8 rows: Price, Free tier, [Key feature 1], [Key feature 2]... Use ✅/❌/partial for quick scanning
4. Write "Choose [X] if..." and "Choose [Y] if..." sections — specific use cases, not generic praise ...

Tips

• The #1 mistake: too generic. If your skill's Instructions say "write a good review," the AI will produce different things every time. Be specific about what "good" means: word count, tone, structure, what to include, what to avoid.
• Test with the worst-case input. If your skill works with "Notion vs Obsidian" but fails with "Descript vs Riverside for podcast recording," your instructions aren't general enough. Test with 2-3 different inputs before publishing.
• The Description is your sales pitch. On LIST, people see the Description in the feed before clicking. "Helps create content" gets zero clicks. "Write comparison posts targeting '[X] vs [Y]' keywords" tells them exactly what they'll get.
• Don't over-instruct. If a step is obvious to any AI model ("use proper grammar"), don't include it. Focus instructions on the parts where AI tends to go wrong — that's where your expertise adds value.