Claude Code Feature Analyzer

Analyze Claude Code features and capabilities in depth, producing actionable guidance on advanced usage, best practices, and real-world application scenarios.

When This Skill Activates

The user provides a Claude Code feature or capability — it could be a feature name like "Streaming Tool Execution", a mechanism like "Hooks", a short question like "MCP 有什么进阶用法？", or a file containing a feature description — and wants to understand how to leverage it effectively in their Claude Code workflow.

How to Work

Step 1: Identify the Feature

Read the user's input carefully. It might be:

• A feature name (e.g., "Hooks", "Streaming Tool Execution", "Subagents")
• A question about a capability (e.g., "Skills 有什么进阶用法和最佳实践？")
• A file path containing a feature description — if so, read the file first

If the feature is ambiguous, briefly clarify with the user. Otherwise, proceed directly.

Step 2: Research the Feature

This step is critical — the analysis must be grounded in how the feature actually works, not generic speculation.

1. Search the Claude Code source code in the current workspace to understand the feature's implementation, configuration options, and internal behavior. Look for relevant source files, types, schemas, and test cases.
2. Search the web for the latest Claude Code documentation, blog posts, and community discussions about this feature. Claude Code evolves rapidly, so up-to-date information matters.
3. Cross-reference what you find in the source code with the documentation to build a complete picture.

Spend real effort here. The quality of the analysis depends entirely on how well you understand the feature's actual capabilities and constraints.

Step 3: Produce the Analysis Report

Generate a structured markdown report. The depth should match the feature's complexity — a simple configuration option gets a focused analysis, while a major capability like Hooks or MCP deserves a thorough deep dive.

Report Template

ALWAYS follow this structure:

# [Feature Name] — Claude Code 进阶指南

## 功能概述
What this feature is and what problem it solves in the Claude Code workflow.
One paragraph, concise and clear.

## 工作原理
How the feature works under the hood. Include:
- The execution flow (what happens when this feature is triggered)
- Key configuration options and their effects
- How it interacts with other Claude Code components

Use diagrams (text-based) or step-by-step flows when the mechanism
involves multiple phases or components.

## 进阶用法
This is the core section. For each advanced usage pattern:
- **Pattern name**: A descriptive title
- **Scenario**: When you'd want to use this pattern
- **How to set it up**: Concrete configuration or commands
- **Example**: A realistic, copy-pasteable example
- **Why it works**: The reasoning behind this approach

Organize from most practical to most creative. Include at least 3-5 patterns.

## 最佳实践
Actionable recommendations for using this feature effectively.
Each practice should explain the reasoning — not just "do X"
but "do X because Y, which prevents Z".

Focus on:
- Configuration best practices
- Performance and efficiency tips
- Security considerations
- Common mistakes and how to avoid them

## 应用场景
Concrete real-world scenarios where this feature shines:
- **Solo development**: How an individual developer benefits
- **Team workflows**: How teams can leverage this feature
- **CI/CD integration**: How to use this in automated pipelines
- **Domain-specific**: Specialized use cases (data science, DevOps, etc.)

For each scenario, describe the problem, the solution using this feature,
and the outcome.

## 注意事项与陷阱
Honest assessment of limitations and gotchas:
- Known limitations or edge cases
- Performance implications
- Compatibility considerations
- Things that look like they should work but don't

## 与其他功能的组合
How this feature combines with other Claude Code capabilities
to create powerful workflows. For example:
- Hooks + Skills for automated workflows
- MCP + Subagents for parallel research
- Context Management + Hooks for session optimization

## 延伸阅读
Pointers to official documentation, relevant source code files,
community resources, and related features to explore next.

Guidelines

• Language: Default to Chinese with English technical terms preserved (e.g., "Hooks", "MCP", "Streaming"). If the user writes in English, respond in English.
• Be concrete: Every recommendation should include a realistic example — a config snippet, a command, or a workflow description. Avoid vague advice like "use it wisely".
• Source-grounded: Reference specific source code files, configuration schemas, or documentation URLs when making claims about how a feature works. This builds trust and lets the user verify.
• Practical over theoretical: Prioritize "here's how to do it" over "here's how it works internally". The user wants to level up their Claude Code usage, not read an architecture doc.

Output

Save the analysis report as <feature-name>-claude-code-guide.md in the current working directory, and display the key highlights in the conversation. If the user specified an output path, use that instead.