Claude Skills: Build Your Own AI Employees

Claude Skills: A Beginner’s Guide to Setup and Optimization

Key Concepts:

• Claude Skills: Customizable instructions that Claude follows consistently across chats, enhancing output quality and consistency.
• Skill Creator: Claude’s built-in tool for creating skills through conversation or direct instruction writing.
• Skill.md: The primary markdown file containing the skill’s workflow and instructions.
• Conversion Copywriting: The practice of writing copy specifically designed to maximize conversion rates (e.g., sales, sign-ups).
• Iteration: The process of refining a skill based on output and feedback to improve performance.
• Offline Mode: A 2-day event for business owners aiming for rapid growth, focusing on AI integration and scaling.

I. Enabling and Creating Skills

The first step to leveraging Claude’s power is enabling the “Skills Preview” feature within settings under “Capabilities.” This allows for the use of repeatable, customizable instructions. Three options are available for skill creation: creating with Claude (conversational onboarding), writing skill instructions directly (for simpler skills), and uploading a pre-built skill (as a .zip file). The recommended approach for beginners is the “create with Claude” option, which guides users through the process.

The skill creation process begins with defining the skill’s functionality. The example provided focuses on creating a skill to review copywriting for apps and websites, specifically aiming to maximize conversion rates. Claude prompts the user with questions to refine the skill’s parameters, including:

• Target Application: Mobile apps with embedded AI or SaaS products.
• Input Method: Screenshots, Figma files, or pasted text.
• Elements for Critique: Headlines, CTAs, value propositions, product descriptions, hero sections, pricing pages, and onboarding flows – essentially all copy elements.
• Desired Review Format: Specific suggestions (before/after examples), scored assessments, and prioritized issue lists.

II. Skill Structure and Functionality – The Conversion Copywriting Example

Following the conversational onboarding, Claude constructs the skill, building the underlying structure. This involves creating markdown (.md) files that define the skill’s logic and knowledge base. Specifically, the example skill includes:

• skill.md: The core workflow document outlining the review process, scoring system, and output template.
• conversion framework.md: Contains established conversion copywriting frameworks like AIDA (Attention, Interest, Desire, Action), PAS (Problem, Agitate, Solve), and BAB (Before, After, Bridge).
• element guidelines.md: Provides best practices for specific copy elements (headlines, CTAs, pricing, etc.).

Claude automatically loads these frameworks and guidelines when prompted to review copy, analyzes each element, assigns a score, prioritizes fixes based on conversion impact (Critical, High, Medium, Low), and provides before/after examples with psychological rationale. The skill is designed to work with various input formats (screenshots, Figma files, pasted text).

III. Skills vs. Projects: A Key Distinction

The speaker emphasizes the difference between Claude Skills and Claude Projects. Projects are context-specific and have a defined beginning and end (e.g., a Christmas campaign). Skills, however, are designed to be persistent and applicable across various contexts, functioning like a dedicated team member with consistent expertise. Skills are ideal for day-to-day operations and core competencies, while projects are suited for temporary, focused initiatives.

IV. Testing and Demonstration

The created skill was tested using screenshots from the CalAI app and website copy. The skill immediately identified issues with generic messaging, lack of specific benefits, and buried value propositions. It provided concrete suggestions for improvement, including alternative headlines ("Track calories in 3 seconds. No manual logging.") and rewrites focusing on transformation ("Lost 17 lbs without counting a single calorie."). The skill also assigned a score to the existing copy (a harsh 3/10 in the example) and clearly outlined the rationale behind its recommendations.

V. Optimizing Skills: Beyond Initial Creation

The speaker highlights a tweet from “boring marketer” emphasizing that effective skills require Claude to think like an expert, not just follow steps. A 10-step process for skill optimization was presented:

1. Define the Problem: Clearly identify the skill’s purpose.
2. Explore: Understand where Claude currently fails without guidance.
3. Research: Deeply investigate the relevant domain.
4. Synthesize: Extract core principles from the research.
5. Draft: Write the initial skill instructions.
6. Self-Critique: Review against quality criteria.
7. Iterate: Fix gaps, gather feedback, and improve.
8. Test: Use the skill in a real-world scenario.
9. Finalize: Codify the skill into an optimal structure.

This iterative process is crucial for transforming a basic skill into a highly effective tool that consistently delivers valuable output. Skills should be continuously refined based on performance and feedback.

VI. Advanced Usage: Cloud Code Integration

While skills can be used directly within the Claude web interface or Mac app, the speaker mentions the potential for even greater control through integration with Cloud Code, allowing for skill utilization within the terminal.

VII. Offline Mode Event Announcement

The speaker briefly promoted an exclusive 2-day event called “Offline Mode” in Fort Lauderdale, Florida (January 23-24). This event is targeted towards business owners generating at least $50,000/month in revenue, focusing on scaling strategies, AI integration, and building multiple products. A link to the event is provided in the video description.

Notable Quote:

“The trick to creating effective skills is to make your AI think like an expert, not just to follow steps.” – Boring Marketer (as quoted by the speaker)

Conclusion:

Claude Skills offer a powerful way to enhance the consistency and quality of Claude’s output. By carefully defining the skill’s purpose, leveraging the Skill Creator tool, and continuously iterating based on performance, users can create customized AI assistants tailored to their specific needs. The key takeaway is that skills are not a one-time setup but an ongoing process of refinement and optimization, ultimately leading to a more valuable and reliable AI experience.