Building Claude Skills
Creates structured Claude Code agent skills from professional expertise. Use when converting methodologies, workflows, or domain knowledge into reusable agent capabilities.
Building Claude Skills
Create professional Claude Code agent skills that capture real expertise in a structured, reusable format.
YAML--- name: analyzing-sales-data description: Analyzes sales performance metrics and identifies trends. Use when reviewing quarterly reports or investigating revenue patterns. --- # Sales Data Analysis Extract insights from sales data using statistical analysis and visualization.
- Data Import - Load CSV/Excel files with pandas
- Clean & Validate - Handle missing values, outliers
- Calculate KPIs - Revenue, conversion rates, growth metrics
- Trend Analysis - Time series patterns, seasonality
- Visualization - Charts showing key insights
- Report Generation - Summary with actionable recommendations
Example 1: Input: Monthly sales data with columns: date, product, revenue, units Output: Growth rate analysis, top performers, seasonal trends chart
- Always validate data quality first
- Focus on actionable insights
- Use clear visualizations
- Don't ignore data quality issues
- Avoid correlation/causation confusion
Progress:
- Extract core methodology from source material
- Identify trigger scenarios for skill usage
- Structure into standardized sections
- Create concrete examples
- Add practical guidelines
Step 1: Analyze Source Material
- Identify the professional domain and specific expertise
- Extract key processes, tools, and methodologies
- Note common use cases and trigger scenarios
Step 2: Structure the Skill
- Write YAML frontmatter with gerund name and trigger-based description
- Lead with Quick Start showing immediate value
- Break methodology into clear workflow steps
- Provide concrete input/output examples
Step 3: Add Practical Value
- Include best practices from the expertise
- Highlight common mistakes to avoid
- Focus on actionable guidance over theory
Example 1:
Input: "I'm a financial analyst who specializes in DCF modeling"
Output: Skill for building-dcf-models with valuation workflow, Excel templates, and sensitivity analysis
Example 2:
Input: "I help companies implement agile methodologies"
Output: Skill for implementing-agile-practices with sprint planning, retrospective formats, and team coaching techniques
Example 3:
Input: GitHub repository with data science utilities
Output: Skill for processing-research-data with analysis pipeline, visualization templates, and statistical methods
- Extract the "how" - Focus on methodology, not just what someone does
- Make it actionable - Each section should enable immediate use
- Use domain language - Include terminology and concepts experts use
- Show real examples - Concrete inputs and outputs beat abstract descriptions
- Keep focused - One skill = one core capability
- Test trigger phrases - Description should match when someone would use this
- Don't create skills for basic tasks everyone knows
- Avoid overly broad skills that try to do everything
- Don't include personal opinions without professional backing
- Skip theoretical background - focus on practical application
- Don't make skills that require extensive setup or rare tools