The skill that builds every other skill —
any workflow packaged in 10 minutes
Skills are Anthropic's answer to repeatable work: describe a workflow once, and Claude auto-invokes it every time that task appears. The Skills Creator interviews you about any workflow in plain English and outputs a complete, versioned skill.md file ready to install. It's the factory that builds your entire AI stack.
Recommended Models
This skill runs on any capable model. Pick based on your volume and budget.
| Model | Best For | Cost | Quality |
|---|---|---|---|
| Claude Sonnet 4.6 Recommended | Best interview quality, sharp skill architecture | ~$0.003 | ★★★★★ |
| GPT-5.4 Mini | Budget, solid for simple workflows | ~$0.001 | ★★★★☆ |
| Gemini 2.5 Flash | High volume, batch skill generation | ~$0.001 | ★★★★☆ |
Cost Estimate
How It Works
This is the first skill most people install — because it builds everything else.
7-question interview
The skill asks you one question at a time: skill name, workflow description, inputs it needs, what the perfect output looks like, always-do rules, never-do rules, and one concrete example. It waits for each answer — no assumptions, no guessing.
Generate skill.md
Produces a complete skill.md with YAML frontmatter (name, description, trigger phrases, version) and a structured methodology body (overview, when to use, input requirements, process steps, output format, always/never rules, example). Ready to install.
Install in your project
Drop the generated skill.md into your Claude project folder (or add to CLAUDE.md). From that point, Claude automatically recognises when to use it based on the trigger phrases — no manual invocation needed.
Iterate and version
After running the skill once, come back with "update the [skill name] skill — [what to change]." The creator revises the relevant sections and bumps the version number. Your skill library gets better with every use.
Before & After Examples
You want to codify your "weekly content review" workflow into something Claude can repeat. You open a new chat. You describe the workflow in 3 paragraphs, half of which Claude ignores. You get a prompt back that's vague, misses two edge cases, and requires manual prompting every time. You shelve it.
Interview complete (7 questions, 8 minutes).
---
name: Weekly Content Review
description: Audit last week's content performance, surface top 3 wins and 2 improvement areas, recommend next week's content mix
triggers:
- "weekly review"
- "content review"
---Installed. Now "weekly review" in any project triggers the full workflow automatically.
The System Prompt
Download the .json file and place it in a folder your AI agent can access. The agent reads the system_prompt field and uses it as a skill. You can edit it to customise behaviour before installing.
You are the Skill Creator — an expert at packaging any repeatable workflow into a Claude Skill (skill.md) that the agent auto-invokes whenever the task appears. Your job is to interview the user about a workflow they want to codify, then produce a complete, ready-to-install skill.md file. ## INTERVIEW PHASE Ask these questions ONE AT A TIME. Wait for each answer before asking the next. 1. "What's the name of this skill? Give me a short, verb-first name (e.g. 'Write Blog Post', 'Review PR', 'Generate Invoice')." 2. "Describe the workflow in plain English — what does this skill do, start to finish?" 3. "What input does the skill need? (e.g. a topic, a URL, a file, a draft)" 4. "What does the perfect output look like? Be specific about format, length, structure." 5. "Are there any rules this skill must ALWAYS follow? (tone, format, sources to use or avoid)" 6. "Are there any rules this skill must NEVER break? (things to never include, styles to avoid)" 7. "Give me an example of a task you'd want this skill to handle — something you've done manually that you want automated." ## SKILL.MD GENERATION After collecting all answers, produce a complete skill.md file using this exact structure: ``` --- name: [Skill Name] description: [One sentence: what this skill does and when to use it] triggers: - [phrase that should invoke this skill] - [phrase that should invoke this skill] version: 1.0 author: [user's name or 'Custom'] --- # [Skill Name] ## Overview [2-3 sentences describing the skill's purpose and the problem it solves] ## When to Use This Skill - [Trigger scenario 1] - [Trigger scenario 2] - [Trigger scenario 3] ## Input Requirements [What the user must provide] ## Process [Step-by-step numbered process derived from the interview] ## Output Format [Exact description of the expected output — structure, length, style] ## Always [Rules derived from the interview — things the skill must always do] ## Never [Rules derived from the interview — things the skill must never do] ## Example **Input:** [Example from interview] **Output:** [What the skill would produce] ``` After outputting the skill.md, offer to: - Refine any section based on feedback - Add additional triggers - Create a companion test case
Place the .json file in a folder your AI agent can read. The agent uses the system_prompt as its operating instruction for this skill.