AI Skill Report Card

Managing Claude Skills

Creates and manages Claude Code agent skills with proper structure and configuration. Use when creating new skills, organizing skill files, or setting up skill frontmatter.

A-88·May 30, 2026·Source: Extension-page
15 / 15

Create a basic skill:

Bash
mkdir -p ~/.claude/skills/my-skill cat > ~/.claude/skills/my-skill/SKILL.md << 'EOF' --- description: Does something specific when triggered ---
Recommendation
Add more specific triggers in the description like 'when creating, organizing, configuring, or debugging Claude skills'
  1. Analyze the input
  2. Process according to requirements
  3. Return formatted output
18 / 20

Example 1: Input: sample input Output: expected result EOF

Recommendation
Include examples of actual skill testing and validation steps with concrete commands
15 / 15

Creating a new skill:

Progress:

  • Create skill directory: mkdir -p ~/.claude/skills/skill-name
  • Write SKILL.md with frontmatter and instructions
  • Test with /skill-name or let Claude auto-invoke
  • Add supporting files if needed
  • Refine based on usage

Skill locations by scope:

  • Personal: ~/.claude/skills/skill-name/ (all projects)
  • Project: .claude/skills/skill-name/ (current project only)
  • Enterprise: Managed settings (organization-wide)
Recommendation
Expand completeness section with troubleshooting common errors and file structure requirements
18 / 20

Example 1: Simple task skill

YAML
--- name: code-review-checklist description: Provides code review checklist for pull requests disable-model-invocation: true ---
  • Code follows style guidelines
  • Tests are included and passing
  • Error handling is appropriate
  • Documentation is updated

**Example 2: Auto-invoked reference skill**
```yaml
---
description: API conventions for this codebase. Use when writing or reviewing API code.
---
  • Use RESTful naming conventions
  • Return consistent error formats with status codes
  • Include request validation
  • Document all endpoints

**Example 3: Skill with dynamic content**
```yaml
---
description: Summarizes git changes and flags risks. Use when reviewing diffs or creating commits.
---

!git diff HEAD

Summarize changes in bullet points and list any risks.

Frontmatter essentials:

  • Always include description with trigger phrases
  • Use disable-model-invocation: true for manual-only skills
  • Set context: fork for complex, isolated tasks
  • Limit allowed-tools for focused workflows

Content guidelines:

  • Keep instructions concise (skills stay in context)
  • Use dynamic injection (!command) for live data
  • Reference supporting files from main SKILL.md
  • Include concrete examples with input/output pairs

Organization:

  • Use kebab-case directory names
  • Group related skills in subdirectories
  • Place frequently-used skills in personal folder
  • Project-specific skills go in .claude/skills/
  • Missing description: Claude won't know when to auto-invoke
  • Verbose instructions: Every line costs tokens across turns
  • Wrong scope: Personal skills override project ones
  • No examples: Abstract instructions confuse execution
  • Forgetting restart: New top-level skill directories need session restart
  • Tool conflicts: disallowed-tools only applies during skill execution
0
Grade A-AI Skill Framework
Scorecard
Criteria Breakdown
Quick Start
15/15
Workflow
15/15
Examples
18/20
Completeness
12/20
Format
15/15
Conciseness
13/15