AI Skill Report Card

Crafting Claude Skills

A-88·Feb 27, 2026·Source: Web
YAML
--- name: crafting-claude-skills description: Creates and improves Claude Code agent skills using expert patterns and best practices. Use when developing new skills or refining existing ones. --- # Crafting Claude Skills
15 / 15
YAML
--- name: analyzing-code-quality description: Analyzes code for maintainability, performance, and best practices. Use when reviewing codebases or preparing code for production. --- # Analyzing Code Quality
Recommendation
Add specific templates or frameworks section with copy-pasteable skill scaffolds for common patterns
15 / 15
Python
# Example analysis output issues = { 'critical': ['No error handling in main function'], 'style': ['Inconsistent naming: mixedCase vs snake_case'], 'performance': ['O(n²) loop could be optimized'] }
Recommendation
Include edge cases like handling skills that are too similar or skills that should be merged
13 / 15
  1. Parse code structure and imports
  2. Check for common antipatterns
  3. Evaluate performance bottlenecks
  4. Generate prioritized recommendations
Recommendation
Provide validation checklist with specific criteria for evaluating skill quality before publishing
13 / 15

Progress:

  • Define skill purpose and trigger phrases
  • Create YAML frontmatter with gerund name
  • Write immediate Quick Start example
  • Structure clear workflow steps
  • Add concrete input/output examples
  • Include actionable best practices
  • List common pitfalls to avoid
  • Validate skill length (~500 lines max)
18 / 20

Example 1: Data Processing Skill Input: "I need to clean CSV files with missing data" Output:

YAML
name: cleaning-csv-data description: Cleans and validates CSV files with missing or malformed data. Use when preprocessing datasets for analysis.

Example 2: Documentation Skill Input: "I write technical documentation for APIs" Output:

YAML
name: writing-api-documentation description: Creates comprehensive API documentation with examples and best practices. Use when documenting REST APIs or SDKs.

Naming Convention:

  • Always use gerund form: processing-images not image-processor
  • Kebab-case only: analyzing-performance not analyzing_performance
  • Be specific: optimizing-sql-queries not optimizing-databases

Description Formula:

  • Action + Object + Context: "Analyzes SQL queries for performance bottlenecks"
  • Always include trigger: "Use when optimizing database performance"
  • Third person only: "Creates..." not "Create..."

Structure Hierarchy:

  1. Quick Start - Code/concrete example first
  2. Workflow - Numbered steps with checkboxes for complex flows
  3. Examples - Always input/output pairs
  4. Best Practices - Actionable guidelines
  5. Common Pitfalls - What NOT to do

Content Guidelines:

  • Default to one recommended approach, not multiple options
  • Assume Claude knows basics (no explaining what JSON is)
  • Include realistic data in examples
  • Keep workflows to 5-8 steps maximum
  • Use code blocks for any technical output

Quality Checks:

  • Can someone use this immediately? (Quick Start test)
  • Are examples copy-pasteable? (Practicality test)
  • Does it solve a specific problem? (Scope test)
  • Is it under 500 lines? (Brevity test)

Naming Mistakes:

  • Using noun forms: code-analyzeranalyzing-code
  • Mixed cases: analyzeCodeanalyzing-code
  • Too generic: helperformatting-json-responses

Description Errors:

  • First/second person: "You can analyze" → "Analyzes"
  • Missing triggers: "Analyzes code" → "Analyzes code for bugs. Use when reviewing pull requests"
  • Vague scope: "Helps with data" → "Cleans CSV files with missing values"

Structure Issues:

  • Starting with explanation instead of example
  • Multiple workflow options instead of one clear path
  • Abstract examples instead of concrete input/output
  • Missing checklists for complex multi-step processes

Content Problems:

  • Over-explaining basic concepts Claude already knows
  • Offering too many alternatives instead of recommended defaults
  • Examples that can't be copy-pasted and used immediately
  • Workflows longer than 8 steps without breaking into sub-skills
0
Grade A-AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
15/15
Workflow
13/15
Examples
18/20
Completeness
13/20
Format
15/15
Conciseness
14/15