Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Progressive disclosure: frontmatter → body → linked files
Hooks
hooks/hooks.json or plugin.json
Event-based: PreToolUse, PostToolUse, SessionStart, etc.
MCP Servers
plugin.json or .mcp.json
"command": "...", "args": [...]
Team Config
.claude/settings.json
extraKnownMarketplaces + plugins.enabled
Hook Event
When It Fires
Common Use
PreToolUse
Before tool execution
Validation, preparation
PostToolUse
After tool execution
Testing, cleanup, linting
SessionStart
Session begins
Logging, initialization
SessionEnd
Session terminates
Cleanup, state save
UserPromptSubmit
After user prompt
Preprocessing, logging
Environment Variable
Purpose
${CLAUDE_PLUGIN_ROOT}
Resolves to plugin's absolute path
${TOOL_INPUT_*}
Access tool input parameters in hooks
When to Use This Skill
Use for advanced plugin features:
Implementing Agent Skills with progressive disclosure
Setting up hook automation (testing, linting, validation)
Configuring MCP server integration
Team plugin distribution via repository configuration
Performance optimization and context efficiency
Migrating legacy plugins to 2025 patterns
For basic plugin creation: see plugin-master skill
🚨 CRITICAL GUIDELINES
Windows File Path Requirements
MANDATORY: Always Use Backslashes on Windows for File Paths
When using Edit or Write tools on Windows, you MUST use backslashes (\) in file paths, NOT forward slashes (/).
Examples:
❌ WRONG: D:/repos/project/file.tsx
✅ CORRECT: D:\repos\project\file.tsx
This applies to:
Edit tool file_path parameter
Write tool file_path parameter
All file operations on Windows systems
Documentation Guidelines
NEVER create new documentation files unless explicitly requested by the user.
Priority: Update existing README.md files rather than creating new documentation
Repository cleanliness: Keep repository root clean - only README.md unless user requests otherwise
Style: Documentation should be concise, direct, and professional - avoid AI-generated tone
User preference: Only create additional .md files when user specifically asks for documentation
Advanced Plugin Features (2025)
Comprehensive guide to cutting-edge Claude Code plugin capabilities introduced in 2025.
Agent Skills
What are Agent Skills?
Skills that Claude autonomously invokes based on task context, enabling dynamic knowledge loading and context-efficient workflows through progressive disclosure architecture.
Key Characteristics (2025)
Automatic Discovery: Skills are discovered from skills/ directory upon plugin installation
Context-Driven: Claude loads Skills only when relevant to the current task
Progressive Disclosure: Three-tier information structure (frontmatter → SKILL.md body → linked files)
Dynamic Loading: Reduces context usage by loading only necessary components
Unbounded Capacity: With filesystem access, agents can bundle effectively unlimited content
---
name: skill-name
description: "Complete [domain] system. PROACTIVELY activate for: (1) [use cases]. Provides: [capabilities]."
license: MIT
---# Skill Title## Overview
High-level summary always loaded
## Core Concepts
Key information organized for quick scanning
## Examples
Concrete usage patterns
## Best Practices
Proven approaches and patterns
Agent Skills Best Practices (2025)
Evaluation-Driven Development:
Start by identifying capability gaps through representative tasks
Observe where agents struggle, then build skills addressing specific shortcomings
Avoid anticipating needs upfront - respond to actual failures
Structural Scalability:
When SKILL.md becomes unwieldy, split content into separate files and reference them
Keep mutually exclusive contexts in distinct paths to reduce token consumption
Code should serve dual purposes: executable tools AND documentation
Split large skills into focused components as needed
Iterative Refinement:
Collaborate with Claude during development
Ask Claude to capture successful approaches into reusable context
Request self-reflection on failure modes to reveal actual information needs
DO:
Use descriptive, action-oriented names
Write comprehensive activation descriptions with numbered use cases
Include concrete examples and code snippets
Organize content with clear headers for scanning
Keep individual skills focused on single domains
DON'T:
Create overly broad skills that cover too many topics
Duplicate content across multiple skills
Skip the frontmatter metadata
Use generic descriptions that don't specify activation scenarios
Context Efficiency Strategy (2025)
Agent Skills achieve unbounded capacity through:
Progressive Disclosure: Three-tier structure (frontmatter loaded at startup → SKILL.md body when determining relevance → additional files only when needed)
Lazy Loading: Only loaded when task matches activation criteria
Filesystem Retrieval: With code execution capabilities, agents retrieve only necessary components instead of loading entire skills
Focused Scope: Each skill covers specific domain/capability
Structured Content: Headers enable Claude to scan efficiently
Optimized Metadata: Name and description fields directly influence triggering accuracy
Example Activation Pattern:
User task: "Deploy to production"
→ Claude scans skill metadata (frontmatter only)
→ Identifies deployment-workflows skill as relevant
→ Loads SKILL.md body to confirm match
→ Retrieves only needed sections/files
→ Unloads when task complete
Result: Effectively unlimited bundled content without context window constraints.
Hooks
What are Hooks?
Automatic triggers that execute actions at specific events during Claude Code's workflow.
Hook Event Types (2025)
Tool Lifecycle:
PreToolUse: Before any tool execution (validation, preparation)
PostToolUse: After tool execution (testing, cleanup, notifications)
Session Lifecycle:
SessionStart: When Claude Code session begins
SessionEnd: When session terminates
PreCompact: Before context compaction (cleanup, state save)
User Interaction:
UserPromptSubmit: After user submits prompt (logging, preprocessing)
{"PostToolUse":[{"matcher":"Write","hooks":[{"type":"command","command":"./scripts/format.sh","env":{"FILE_PATH":"${TOOL_INPUT_FILE_PATH}"}}]}],"SessionStart":[{"hooks":[{"type":"command","command":"echo 'Session started at $(date)' >> session.log"}]}]}
Hook Matchers
Tool Matchers:
Match specific tools using regex patterns:
Write - File write operations
Edit - File edit operations
Write|Edit - Either write or edit
Bash - Shell command execution
.* - Any tool (use sparingly)
Matcher Best Practices:
Be specific to avoid unnecessary executions
Use | for multiple tools
Test matchers thoroughly
Document why each matcher is used
Common Hook Patterns
Automated Testing:
{"PostToolUse":[{"matcher":"Write|Edit","hooks":[{"type":"command","command":"${CLAUDE_PLUGIN_ROOT}/scripts/run-tests.sh","description":"Run tests after code changes"}]}]}
## Claude Code Setup
This repository uses Claude Code plugins for standardized workflows.
### First Time Setup1. Install Claude Code
2. Clone this repository
3. Open in Claude Code
4. Trust this repository folder when prompted
Plugins will be installed automatically.
Security Considerations:
Only trust repositories from known sources
Review settings.json before trusting
Use organizational repositories for internal plugins
Document why each plugin is required
Environment Variables
Standard Variables (2025)
${CLAUDE_PLUGIN_ROOT}:
Resolves to plugin's absolute installation path.
