Skill Creator

Create effective skills that extend Claude's capabilities.

Core Principle

Claude is already very smart. Only add context Claude doesn't already have. Prefer concise examples over verbose explanations.

Official Documentation

•Agent Skills Specification: https://agentskills.io/specification
•Building Skills Guide: https://resources.anthropic.com/hubfs/The-Complete-Guide-to-Building-Skill-for-Claude.pdf
•Claude Code Skills: https://code.claude.com/docs/en/skills

Progressive Disclosure (3-Level Loading)

Level	What	Token Budget	When Loaded
1. Metadata	name + description	~100 tokens	Always (all skills)
2. Body	SKILL.md content	<5000 tokens	When skill triggers
3. Resources	references/, scripts/, assets/	Unlimited	On-demand by Claude

Key insight: Keep SKILL.md lean. Move detailed content to references/.

Skill Structure

code

skill-name/
├── SKILL.md           # Required: frontmatter + instructions (<500 lines)
├── references/        # Optional: documentation (on-demand loading)
├── scripts/           # Optional: executable code
└── assets/            # Optional: templates, images

DO NOT create: README.md, CHANGELOG.md, or other auxiliary files.

SKILL.md Format

Frontmatter (Required)

yaml

---
name: kebab-case-name
description: [WHAT it does]. [WHEN to use]. Triggers on "english1", "english2", "日本語トリガー".
model: haiku|sonnet|opus
context: fork          # Required for Code Gen/Workflow, optional for Guidance
agent: general-purpose
allowed-tools: Read, Glob, Grep, Bash(specific:*)
---

Triggers must include both English AND Japanese for accessibility.

Field Constraints (Agent Skills Spec):

Field	Required	Constraints
name	Yes	1-64 chars, lowercase + hyphens, must match folder name
description	Yes	1-1024 chars, WHAT + WHEN + Triggers

Model Selection (with numeric criteria):

Model	Use When	Criteria
haiku	Read-only guidance	0 file writes, 0 external calls
sonnet	Code generation	1-3 steps, ≤2 external services
opus	Complex workflows	4+ steps OR 3+ services OR MCP tools

allowed-tools: Use specific patterns like Bash(git:*), Bash(gh:*), not generic Bash.

Invocation Control

yaml

disable-model-invocation: true  # User must invoke manually with /skill-name
user-invocable: false           # Only Claude can invoke (hidden from / menu)

When to use disable-model-invocation: true:

Pattern	Example	Reason
Setup/Init	repository-setup, docusaurus-setup	One-time setup should be intentional
Release/Deploy	swift-release	Side effects, requires user confirmation
Config changes	automation-config	Modifies project configuration
Complex creation	swift-mcp-server	Large-scale changes, user should control timing

When to keep default (auto-invocation allowed):

Pattern	Example	Reason
Daily tasks	commit-message, vigilare-task	Frequent use, "コミットメッセージ" triggers naturally
Code assistance	swift-development, documentation	Helps during normal development flow
Workflow shortcuts	github-workflow	"Issue作って" should just work

Body

Concise instructions only. Move detailed content to references/.

When Invoked

Step 0: Research (if needed)

When to research:

•Creating a new type of skill (no similar existing skill)
•User asks for "best practices" or "latest patterns"
•Unfamiliar domain or tool integration

Research targets (parallel):

Target	Method
Agent Skills Spec	WebFetch agentskills.io/specification
Existing skills	Task(Explore) in skills/ directory
Other marketplaces	WebSearch "claude code skills marketplace"
Similar tools	WebSearch "{tool-name} claude skill"

Skip research when:

•Simple modification to existing skill
•Pattern already exists in this repository
•User provides clear, complete requirements

Step 1: Gather Requirements

Essential:

•What should the skill do?
•Example usage scenarios?
•What tools/commands are needed?

Invocation & scope:

•Should it auto-invoke, or manual only? → disable-model-invocation
•Who uses it? JP users, EN users, or both? → Triggers in description
•Does it have side effects (deploy, release, config)? → manual only

Step 2: Determine Skill Type

Type	context: fork	model	Criteria	Example
Guidance	Optional	haiku	0 writes, 0 services	documentation
Code Gen	Yes	sonnet	1-3 steps, ≤2 services	swift-development
Workflow	Yes	opus	4+ steps OR 3+ services OR MCP	vigilare-task

context: fork: Use for context isolation. Required for Code Gen/Workflow. Optional for Guidance when you want to avoid polluting the main conversation context.

Quick check: Does it use MCP tools? → opus. Does it write files? → sonnet+. Read-only? → haiku.

Step 3: Plan Structure

Identify:

•Required allowed-tools with specific patterns
•Reference files needed
•Scripts or assets (if any)

Step 4: Create Skill

•Create directory: skills/{skill-name}/
•Create SKILL.md with proper frontmatter
•Create references/ if needed
•Add to .claude-plugin/marketplace.json

Step 5: Validate

Agent Skills Spec Compliance:

• name: 1-64 chars, lowercase + hyphens only, matches folder
• name: no leading/trailing hyphens, no consecutive hyphens (--)
• description: 1-1024 chars, non-empty

Labee Standards:

• description has WHAT + WHEN + Triggers (JP & EN)
• allowed-tools uses specific patterns (not generic Bash)
• model selected by numeric criteria (see references/output-patterns.md)
• context: fork for workflow skills (sonnet/opus)
• disable-model-invocation: true for setup/release/config skills

Context Efficiency:

• SKILL.md body under 500 lines
• No redundant explanations (Claude is smart)
• Detailed content moved to references/
• No non-existent reference files listed

Reference Files

Loading behavior:

•Standard files: Claude loads on-demand when the task requires them
•Auto-load files (_ prefix): Always loaded when skill triggers

File	Load When
references/workflows.md	Creating skills with 3+ steps or conditional logic
references/output-patterns.md	Defining description format, allowed-tools patterns, or model selection
references/resource-patterns.md	Deciding scripts/ vs references/ vs assets/, implementing scripts
references/autonomous-refinement-loop.md	Creating skills that need self-correction (verify→fix→retry loops)

Session Learning Scripts

For Autonomous Refinement Loop enabled skills:

Script	Purpose
`scripts/init_session.ts`	Initialize session learning file
`scripts/log_iteration.ts`	Log each verification attempt
`scripts/finalize_session.ts`	Finalize session with learnings

Auto-load guidelines:

•Use _ prefix for core rules that apply to ALL invocations (e.g., _core-rules.md)
•Keep auto-load files small (<200 lines) to preserve token budget
•Examples: documentation/_core-rules.md, research/_source-patterns.md