Skip to content

Skill Construction Roadmap

Organize the classification system, evaluation criteria, dependency relationships, and construction plans for Skills.

日本語

About This Document

This document formalizes Discussion #20 (Skill Construction Strategy Map) as an official project document.

While MCPs handle "real-time access to external data," Skills handle "static knowledge and judgment criteria" (see 03-architecture.md). For the MCP construction strategy, see mcp-roadmap.md. For composition patterns, see composition-patterns.md.

5 Skill Types

All Skills are "static knowledge," but the structure differs depending on the type of knowledge.

TypeForm of AI InstructionApproximate Volume
Principle Type"Determine whether this violates the principle"Short (1-3 pages)
Pattern Type"Select an applicable pattern for this situation"Long (proportional to number of patterns)
Methodology Type"Execute following this process"Medium
Checklist Type"Check all items and report the results"Short
Guideline Type"Improve according to this guideline"Medium

5-Axis Skill Evaluation

Skill-specific evaluation criteria corresponding to MCP's 5 axes (Governing Body, Openness, Machine Readability, Structural Clarity, Practical Use Cases).

AxisMeaning
Knowledge StabilityHow frequently it is updated (Stable ◎ to Volatile △)
Scope ClarityWhether the boundary with other Skills is clear
Trigger ClarityWhether it is possible to determine when to use it
Standalone CompletenessWhether it can deliver value alone, or requires other Skills
VerifiabilityWhether the effect can be objectively measured

Individual Skill Evaluations

Foundation Skills

SkillTypeStabilityScopeTriggerStandaloneVerifiability
SOLIDPrinciple
Clean CodeGuideline
TDDMethodology

Design Skills

SkillTypeStabilityScopeTriggerStandaloneVerifiability
Design PatternsPattern
Clean ArchitecturePrinciple
DDDMethodology

Testing Skills

SkillTypeStabilityScopeTriggerStandaloneVerifiability
Test PatternsPattern
BDDMethodology
Coverage StrategyChecklist

Infrastructure & Architecture Skills

SkillTypeStabilityScopeTriggerStandaloneVerifiability
12 Factor AppPrinciple
MicroservicesPattern
EIPPattern
IaC PatternsPattern

Quality Improvement Skills

SkillTypeStabilityScopeTriggerStandaloneVerifiability
RefactoringPattern

Skill Dependency Relationships

Skills have a clear dependency hierarchy. While MCPs can be built independently, Skills require a build order that considers dependencies.

Trigger Contexts — When to Use

Skills tend to have ambiguous "when to trigger" conditions. Here we organize them by context-based Skill sets.

ContextTriggered SkillsCombination Purpose
Writing CodeClean CodeConstant reference for naming and function design standards
Code ReviewSOLID + Clean Code + RefactoringDetect principle violations → Identify smells → Suggest improvements
Architecture DesignDDD + Clean Architecture + 12 FactorDomain modeling → Layer design → Cloud readiness
Test DesignTDD + BDD + Test Patterns + CoverageProcess → Scenarios → Pattern application → Coverage assessment
RefactoringRefactoring + Design Patterns + SOLIDDetect smells → Apply patterns → Verify principle compliance
Infrastructure Design12 Factor + Microservices + IaC + EIPCloud-native → Decomposition → Automation → Integration

For details on multi-Skill simultaneous trigger patterns, see composition-patterns.md.

Challenges of "Ambiguity" and Countermeasures

Skill design has unique challenges that MCPs do not face.

Challenge 1: Scope Overlap

Is "SRP (Single Responsibility Principle)" part of the SOLID Skill or the Clean Code Skill?

Countermeasure: Place the definition of the principle in the SOLID Skill, and reference it as an application example in Clean Code. Establish a "primary source of definition" among Skills.

Challenge 2: Granularity Decisions

Should DDD be one Skill? Or should it be split into Strategic Patterns and Tactical Patterns?

Countermeasure: If the trigger context is the same, keep it as one. Start with one and decide on separation through operational experience.

Challenge 3: Quality Measurement

How do you measure "Is this a good Skill?"

Countermeasure: Measure using the following indirect metrics.

MetricMeasurement Method
AI Judgment AccuracyWhether review feedback became more accurate after Skill application
Trigger AppropriatenessWhether it avoids triggering in unnecessary situations and triggers in necessary ones
Context EfficiencyThe effect gained relative to token consumption from Skill references

Priority Roadmap

Build order considering dependencies and practical relevance.

Released under the MIT License.