Skip to content

Skill Showcase

Real-world examples of Skills in production. For those who want to see implementation, not just concepts.

About This Document

After learning concepts in What is Skills and design methods in Creating Skills, this page presents three production Skill implementations. Each demonstrates a different design approach, MCP integration depth, and distribution format, providing a practical reference for Skill design patterns.

Examples Overview

SkillPatternMCP IntegrationDistribution
deepl-glossary-translationMCP Orchestrationpdf-spec-mcp + DeepL MCP.claude/skills/
code-review-skillChecklist-basedNone (static knowledge only).claude/skills/ / Project Knowledge
spec-compliance-skillsData-driven ValidationNone (pre-parsed data)Cowork Plugin

deepl-glossary-translation

Repository: shuji-bonji/deepl-glossary-translation

Overview

A Skill that integrates pdf-spec-mcp and DeepL MCP Server to translate the PDF specification (ISO 32000-2) into Japanese with consistent terminology. MCP provides "what can be done" while the Skill defines "how to do it" — a quintessential example of the three-layer architecture.

Architecture

Design Highlights

Multi-MCP Orchestration: Chains pdf-spec-mcp (data extraction) → DeepL MCP (translation) → xcomet-mcp (quality evaluation). The Skill itself has no tools — it defines the "usage patterns" for existing MCPs.

5-Step Workflow Definition: Term extraction → Classification → Glossary registration → Translation → Quality verification. Clear steps in Markdown that agents follow autonomously.

External Script Augmentation: Glossary registration requires direct DeepL API calls, so shell scripts are bundled in scripts/. This pattern complements Skill's static knowledge with executable scripts.

Classification

Maps to Pattern 4: MCPs + Skills (Full Integration) in Composition Patterns.


code-review-skill

Repository: shuji-bonji/code-review-skill

Overview

A Skill for conducting code reviews on TypeScript / MCP Server projects. Defines 10-category evaluation criteria and a 3-tier priority framework for consistent review quality.

10-Category Checklist

CategoryExample Checks
Bug DetectionPlaceholder detection, null checks, async handling
User MessagesLanguage consistency, error actionability
Code Suppressionseslint-disable / @ts-ignore necessity verification
HardcodingMagic numbers, embedded config values
PerformanceN+1 problems, cache gaps, sequential async calls
RedundancyCode duplication, unused variables, dead code
Type SafetyEliminate any types, validate external data with Zod
MaintainabilityResponsibility separation, dependency direction
Code StructureFile size >300 lines, function >50 lines, circular dependencies
Project-specificMCP: tool descriptions, Zod .strict() / Angular: component separation

Priority Framework

LevelTargets
🔴 HighRuntime bugs, security issues, performance degradation
🟡 MediumBug-prone structures, weakened type safety
🟢 LowReadability improvements, minor refactoring

Design Highlights

Pure Skill (no MCP required): No external API calls needed. Works by combining with the agent's built-in code analysis capabilities. The simplest form of a Skill.

Split Placement Design: Recommends different detail levels for Claude.ai (Project Knowledge) vs Claude Code (.claude/). A PLACEMENT-GUIDE.md addresses context window constraints.

Classification

A Skills-only pattern in the three-layer architecture. Demonstrates that Skills can be fully effective without MCP.


spec-compliance-skills

Repository: shuji-bonji/spec-compliance-skills

Overview

A Cowork Plugin that validates document compliance against W3C/IETF specifications. Ships with pre-parsed normative requirements (MUST/SHOULD/MAY) as embedded data, generating structured validation reports.

Included Skills

epub-quality Validation Flow

Design Highlights

Data-driven Validation: Pre-parses normative requirements from specifications into JSON, bundled within the Skill. Enables validation without real-time MCP server access.

Skill → Plugin Evolution: Functions as a standalone Skill but distributes as a Cowork Plugin (.plugin) for easier installation and updates. An example of Skill distribution evolution.

Incremental Expansion: Released epub-quality first, with pdf-quality and rfc-compliance added progressively. A pattern for bundling multiple Skills in a single Plugin.

Classification

A Skills-only (data-embedded) pattern in the three-layer architecture. Instead of real-time data from MCP, it embeds pre-parsed data within the Skill.


Skill Design Space

Skills start with the simple definition of "knowledge written in Markdown," but in practice they encompass diverse design patterns: bundled data files, companion scripts, MCP integration definitions, and Plugin distribution. The key insight is that the essence of a Skill (providing knowledge to agents) remains constant across all patterns.

Released under the MIT License.