doc-fetcher（文档获取器）

Purpose

The doc-fetcher skill provides comprehensive capabilities for fetching library and framework documentation from multiple sources using MCP integrations. This skill helps the Documentation Researcher agent retrieve up-to-date, version-specific documentation that enables informed implementation decisions and adherence to library best practices.

This skill emphasizes:

• Latest Documentation: Always fetch current, version-specific documentation
• Multiple Sources: Leverage both context7-mcp (deep context) and fetch-mcp (web content)
• Comprehensive Coverage: Retrieve API references, examples, guides, and best practices
• Version Awareness: Track version compatibility and breaking changes
• Efficient Retrieval: Optimize MCP usage for token efficiency and accuracy

The doc-fetcher skill ensures that implementation guidance is based on authoritative, current documentation from official sources.

When to Use

This skill auto-activates when the agent describes:

• "Fetch documentation for..."
• "Retrieve API reference for..."
• "Get library documentation..."
• "Access documentation for..."
• "Find documentation about..."
• "Retrieve latest docs for..."
• "Look up documentation..."
• "Fetch API docs from..."

Provided Capabilities

1. Context7-MCP Documentation Retrieval

What it provides:

• Deep documentation retrieval with semantic search
• Version-specific API references
• Code examples from documentation
• Library-specific patterns and conventions
• Integration guidance

Context7 Workflow:

# Step 1: Resolve library name to context7 ID
library_id = invoke_mcp(
    "context7-mcp",
    tool="resolve-library-id",
    params={
        "libraryName": "fastapi"  # or "react", "django", etc.
    }
)

# Step 2: Fetch comprehensive documentation
docs = invoke_mcp(
    "context7-mcp",
    tool="get-library-docs",
    params={
        "context7CompatibleLibraryID": library_id["library_id"],  # e.g., "/tiangolo/fastapi"
        "topic": "API routing and dependency injection",  # Focus area
        "tokens": 3000  # Amount of documentation to retrieve
    }
)

# Result contains:
# - documentation: Markdown-formatted docs
# - version: Library version
# - examples: Code examples
# - metadata: Additional context

Context7 Best Practices:

• Use specific topics to focus documentation retrieval
• Start with 2000-3000 tokens for comprehensive coverage
• Adjust token count based on complexity
• Combine multiple focused queries for complex features

2. Fetch-MCP Web Content Retrieval

What it provides:

• Official documentation page fetching
• GitHub README and Wiki retrieval
• Community resource access
• Tutorial and guide retrieval
• Changelog and migration guide access

Fetch-MCP Workflow:

# Fetch official documentation page
official_docs = invoke_mcp(
    "fetch-mcp",
    tool="fetch",
    params={
        "url": "https://fastapi.tiangolo.com/tutorial/first-steps/",
        "prompt": "Extract quick start guide, installation steps, and first API example"
    }
)

# Fetch GitHub README
github_readme = invoke_mcp(
    "fetch-mcp",
    tool="fetch",
    params={
        "url": "https://github.com/tiangolo/fastapi/blob/master/README.md",
        "prompt": "Extract key features, installation, and basic usage examples"
    }
)

# Result contains:
# - Extracted content focused on the prompt
# - Markdown-formatted for easy parsing
# - Cleaned and processed for relevance

Fetch-MCP Best Practices:

• Use specific prompts to extract relevant content
• Prefer official documentation URLs over third-party
• Fetch READMEs for overview and quick start
• Retrieve changelogs for version migration info

3. Multi-Source Documentation Strategy

What it provides:

• Combined documentation from multiple sources
• Cross-reference validation
• Comprehensive coverage
• Authoritative source prioritization

Multi-Source Workflow:

documentation_sources = {
    "primary": {
        # Context7: Deep, comprehensive docs
        "context7": fetch_via_context7(library_name, topic),

        # Official docs: Quick start and guides
        "official": fetch_via_fetch_mcp(official_docs_url),
    },
    "supplementary": {
        # GitHub: Latest examples and README
        "github": fetch_via_fetch_mcp(github_url),

        # Migration guides (if version upgrade)
        "migration": fetch_via_fetch_mcp(migration_guide_url) if needs_migration else None,
    }
}

# Synthesize documentation from multiple sources
synthesized_docs = synthesize_documentation(documentation_sources)

Source Prioritization:

1. Official Documentation (highest priority)
2. Context7 Documentation (comprehensive reference)
3. GitHub Repository (latest examples)
4. Community Resources (supplementary)

4. Version-Specific Retrieval

What it provides:

• Version compatibility checking
• Breaking change identification
• Migration guidance
• Deprecated feature detection

Version Handling:

# Specify version in context7 (if supported)
docs_v2 = invoke_mcp(
    "context7-mcp",
    tool="get-library-docs",
    params={
        "context7CompatibleLibraryID": "/tiangolo/fastapi/v0.100.0",  # Version-specific
        "topic": "API routing",
        "tokens": 2000
    }
)

# Fetch version-specific changelog
changelog = invoke_mcp(
    "fetch-mcp",
    tool="fetch",
    params={
        "url": "https://github.com/tiangolo/fastapi/blob/master/CHANGELOG.md",
        "prompt": "Extract changes between version 0.95.0 and 0.100.0, focusing on breaking changes"
    }
)

# Compare versions and identify migration needs
migration_notes = analyze_version_changes(changelog)

5. Code Example Extraction

What it provides:

• Working code examples
• Integration patterns
• Configuration examples
• Test examples

Example Extraction:

# Extract examples from context7 docs
examples = []
for code_block in docs["examples"]:
    examples.append({
        "code": code_block["code"],
        "language": code_block["language"],
        "description": code_block["description"],
        "category": categorize_example(code_block)
    })

# Extract examples from official docs
official_examples = extract_code_blocks(official_docs["content"], language="python")

# Combine and deduplicate
all_examples = deduplicate_examples(examples + official_examples)

6. Documentation Caching

What it provides:

• Reduced MCP calls
• Faster subsequent retrievals
• Token usage optimization
• Consistent documentation state

Caching Strategy:

# Check cache before fetching
cache_key = f"{library_name}:{version}:{topic_hash}"

if cache_key in documentation_cache:
    return documentation_cache[cache_key]

# Fetch and cache
docs = fetch_documentation(library_name, version, topic)
documentation_cache[cache_key] = docs
documentation_cache[cache_key]["cached_at"] = datetime.utcnow()

return docs

Cache Invalidation:

• Expire after 24 hours
• Clear on version change
• Manual refresh option

Usage Guide

Step 1: Identify Documentation Needs

Analysis doc → Extract libraries → Identify topics → Prioritize sources

Step 2: Resolve Library IDs (context7)

Library name → context7 resolve-library-id → Library ID

Step 3: Fetch Context7 Documentation

Library ID + Topic → get-library-docs → Comprehensive docs

Step 4: Fetch Web Resources (fetch-mcp)

URLs + Prompts → fetch → Supplementary docs

Step 5: Extract Examples

Documentation → Code blocks → Categorized examples

Step 6: Synthesize Documentation

Multiple sources → Prioritize → Combine → Structured output

Best Practices

1. Use Context7 for Depth

- Primary source for API references - Comprehensive coverage of library features - Semantic search capabilities - Version-specific support

1. Use Fetch-MCP for Breadth

- Official quick start guides - GitHub examples and READMEs - Migration guides and changelogs - Community tutorials (verified sources)

1. Focus Documentation Retrieval

- Use specific topics in context7 - Use targeted prompts in fetch-mcp - Avoid generic "get all documentation" - Retrieve only what's needed for feature

1. Version Awareness

- Always specify version requirements - Check for breaking changes - Document version compatibility - Provide migration notes if needed

1. Token Optimization

- Start with 2000-3000 tokens - Adjust based on complexity - Multiple focused queries > one large query - Cache frequently accessed docs

1. Source Validation

- Prefer official documentation - Verify URL authenticity - Check documentation date - Cross-reference when uncertain

Resources

doc-sources.md

Curated list of documentation sources:

• Official documentation URLs by framework
• GitHub repository locations
• Community resource repositories
• API reference locations
• Tutorial and guide sources
• Changelog and release note locations

fetching-strategies.md

Documentation fetching strategies:

• When to use context7 vs fetch-mcp
• Token optimization techniques
• Version-specific retrieval
• Caching strategies
• Error handling patterns
• Multi-source synthesis
• Example extraction methods

Example Usage

Input (from Documentation Researcher agent):

"Fetch documentation for FastAPI framework focusing on API routing, dependency injection, and Pydantic integration."

Output (doc-fetcher skill provides):

# Comprehensive documentation retrieval

# 1. Context7 Documentation
fastapi_docs = {
    "library_id": "/tiangolo/fastapi",
    "version": "0.100.0",
    "documentation": """
    # FastAPI API Routing

    FastAPI provides a powerful routing system based on Python type hints...

    ## Dependency Injection

    FastAPI's dependency injection system allows you to declare dependencies...

    ## Pydantic Integration

    FastAPI uses Pydantic models for request validation...
    """,
    "examples": [
        {
            "title": "Basic API with dependency injection",
            "code": """
from fastapi import FastAPI, Depends

app = FastAPI()

def get_query_param(q: str = None):
    return {"q": q}

@app.get("/items/")
async def read_items(commons: dict = Depends(get_query_param)):
    return commons
            """,
            "language": "python"
        }
    ]
}

# 2. Official Documentation (fetch-mcp)
official_docs = {
    "url": "https://fastapi.tiangolo.com/tutorial/",
    "content": """
    # First Steps

    Create a file `main.py` with:

from fastapi import FastAPI

app = FastAPI()

@app.get("/") async def root(): return {"message": "Hello World"}

    Run the server with: `uvicorn main:app --reload`
    """
}

# 3. GitHub README (fetch-mcp)
github_readme = {
    "url": "https://github.com/tiangolo/fastapi",
    "content": """
    # FastAPI

    FastAPI framework, high performance, easy to learn, fast to code, ready for production

    ## Key features:
    - Fast: Very high performance, on par with NodeJS and Go
    - Fast to code: Increase the speed to develop features by about 200% to 300%
    - Fewer bugs: Reduce about 40% of human errors
    - Intuitive: Great editor support
    - Easy: Designed to be easy to use and learn
    - Short: Minimize code duplication
    - Robust: Get production-ready code
    - Standards-based: Based on OpenAPI and JSON Schema
    """
}

# 4. Synthesized Output
{
    "library": "fastapi",
    "version": "0.100.0",
    "sources": {
        "context7": "Primary documentation source",
        "official": "Quick start and tutorials",
        "github": "Overview and features"
    },
    "api_routing": {
        "overview": "FastAPI provides decorator-based routing...",
        "examples": [...],
        "best_practices": [...]
    },
    "dependency_injection": {
        "overview": "Dependency injection via Depends()...",
        "examples": [...],
        "best_practices": [...]
    },
    "pydantic_integration": {
        "overview": "Pydantic models for validation...",
        "examples": [...],
        "best_practices": [...]
    },
    "version_notes": "Compatible with Pydantic v2.x"
}

Integration

Used By:

• @documentation-researcher (Primary) - Phase 2 sub-agent for documentation research

Integrates With:

• doc-analyzer skill - Fetched documentation is analyzed for patterns and best practices
• context7-mcp - Primary documentation retrieval mechanism
• fetch-mcp - Supplementary web content retrieval

Workflow Position:

1. Analysis Specialist identifies technical stack requirements
2. Documentation Researcher receives analysis
3. doc-fetcher skill retrieves documentation (Step 3-4)
4. doc-analyzer skill analyzes documentation (Step 5-6)
5. Results synthesized into documentation summary
6. Design Orchestrator includes in PRP

Version: 2.0.0 Auto-Activation: Yes Phase: 2 - Design & Planning Created: 2025-10-29