技能 skill-developer

🛠️

skill-developer

Name: skill-developer
Author: diet103

安全 ⚙️ 外部命令🌐 網路存取📁 檔案系統存取

Create Claude Code Skills

也可從以下取得: 0Chan-smc,Dimon94,0Chan-smc,DojoCodingLabs

Master the art of building and managing Claude Code skills with this comprehensive guide. Learn to configure auto-activation triggers, implement guardrails, set up hooks, and follow Anthropic best practices for scalable skill development.

支援: Claude Code(CC)

📊 70 充足

下載技能 ZIP

在 Claude 中上傳

前往設定 → 功能 → 技能 → 上傳技能

開啟並開始使用

測試它

正在使用「skill-developer」。 How do I configure a skill to activate when someone mentions creating a database table?

預期結果:

Configure keywords: ["database", "table", "schema"]
Add intent pattern: "(create|add|modify).*?(table|database|schema)"
Set enforcement level to suggest for advisory guidance

正在使用「skill-developer」。 What is the 500-line rule and how do I implement progressive disclosure?

預期結果:

The 500-line rule limits skill complexity to maintain readability
Progressive disclosure means showing basic info first, detailed docs on demand
Split complex skills into base skill + specialized sub-skills

正在使用「skill-developer」。 Help me set up a UserPromptSubmit hook for skill suggestions

預期結果:

Create .claude/hooks/skill-activation-prompt.ts
Register hook in .claude/settings.json with UserPromptSubmit type
Hook reads prompt, matches against skill-rules.json, outputs suggestions to stdout

安全審計

安全

v4 • 1/17/2026

All 342 static findings are FALSE POSITIVES. This skill contains only markdown documentation files describing Claude Code skill architecture. The scanner misinterpreted markdown backticks for code syntax highlighting as Ruby/shell command execution, configuration references as filesystem access, and troubleshooting commands as system reconnaissance. No executable code, network operations, or credential access exists in this skill.

已掃描檔案

2,430

分析行數

發現項

審計總數

風險因素

⚙️ 外部命令 (293)

🌐 網路存取 (5)

skill-report.json:158 skill-report.json:6 TRIGGER_TYPES.md:86 TRIGGER_TYPES.md:277 TROUBLESHOOTING.md:46

📁 檔案系統存取 (2)

skill-report.json:6 TROUBLESHOOTING.md:375

審計者: claude 查看審計歷史 →

品質評分

架構

100

可維護性

內容

社群

100

安全

規範符合性

你能建構什麼

Learn skill fundamentals

Understand skill structure, YAML frontmatter, and trigger types for your first skill.

Design guardrail skills

Create enforcement skills that block critical errors before code commits.

Configure hook systems

Set up UserPromptSubmit and PreToolUse hooks for automated workflow guidance.

試試這些提示

Create My First Skill

Help me create a new Claude Code skill for managing database schemas. Include the SKILL.md structure and skill-rules.json configuration.

Add Trigger Patterns

Show me how to add intent patterns and file path triggers to my skill-rules.json for detecting React component creation requests.

Implement Guardrail

Create a guardrail skill that blocks commits without proper error handling. Include enforcement level and block message configuration.

Debug Activation

My skill is not activating when it should. Walk me through the debugging steps in TROUBLESHOOTING.md to identify the issue.

最佳實務

Use specific keywords and test intent patterns with real user prompts before deploying skills
Implement session tracking to avoid repeating suggestions within the same conversation
Keep skills focused with the 500-line rule and use progressive disclosure for complex topics

避免

Avoid overly generic keywords that trigger on unrelated prompts (like 'create' alone)
Do not use block enforcement for non-critical guidance - use suggest or warn instead
Avoid creating monolithic skills - split into focused sub-skills with clear boundaries

常見問題

What file formats do skills support?

Skills use SKILL.md for documentation and skill-rules.json for configuration. Hooks are TypeScript files in .claude/hooks/.

How do I test my skill triggers?

Test skill triggers using TROUBLESHOOTING.md commands to validate JSON syntax and verify intent patterns. Use the UserPromptSubmit hook test command with your actual prompts, and validate regex patterns at regex101.com before deployment to ensure accurate matching without false positives.

Can skills communicate with each other?

Skills are independent but can reference each other through documentation. The skill-rules.json configuration handles activation coordination.

What enforcement levels are available?

Block stops execution, suggest provides advisory prompts, and warn issues notices without blocking. Use block for critical guardrails only.

How do skip conditions work?

Skip conditions in skill-rules.json can skip based on session skill usage, file markers like @skip-validation, or environment variable overrides.

Where are skills stored?

Skills live in .claude/skills/{skill-name}/ with SKILL.md and skill-rules.json. Hooks are in .claude/hooks/ directory.