AgentSkillsCN

web-research-documenter

当用户请求研究任何需要网络搜索、来源分析与结构化文档的主题时，应使用此技能。它提供系统化的工作流，从多个网络来源收集信息，合成结论为全面文档，并保存结果到用户指定的文件路径。这是一个通用的研究技能，适用于所有领域与语言。

SKILL.md
--- frontmatter
name: web-research-documenter
description: This skill should be used when users request research on any topic that requires web search, source analysis, and structured documentation saved to a file. It provides a systematic workflow for gathering information from multiple web sources, synthesizing findings into a comprehensive document, and saving results to a user-specified file path. This is a general-purpose research skill that works across all domains and languages.

Web Research Documenter

Overview

This skill provides a systematic workflow for researching any topic by conducting web searches, analyzing authoritative sources, synthesizing information into well-structured documents, and saving results to user-specified file paths. Unlike domain-specific research skills, this is a general-purpose skill that works across all topics and supports multiple languages.

When to Use This Skill

Use this skill when users request:

  • •Research on any topic with file output
  • •Information gathering from web sources
  • •Structured documentation of findings
  • •Topic summaries or reports
  • •Multi-source analysis and synthesis

Trigger patterns include:

  • •"[Topic]을/를 검색하고 파일로 저장해줘" (Korean)
  • •"Research [topic] and save to file"
  • •"Search for [topic] and create a report"
  • •"Find information about [topic] and document it"
  • •"Investigate [topic] and save the summary"

Key indicators:

  • •Request involves web search/research
  • •Request specifies file output or saving results
  • •Request requires synthesis from multiple sources
  • •Request asks for structured documentation

Workflow

Follow this systematic four-phase workflow:

Phase 1: Initial Web Search

Conduct broad web search to identify key themes and authoritative sources.

Steps:

  1. •Formulate search query based on user's topic and language
  2. •Use web_search tool with appropriate max_results (typically 8-12)
  3. •Scan results to identify:
    • •Major themes and key information
    • •Authoritative sources
    • •Specific data points or statistics
    • •Key terminology and concepts

Search Query Guidelines:

  • •Match the user's language (if user asks in Korean, search in Korean)
  • •Include temporal indicators when relevant: "2024", "2025", "latest", "recent", "최신"
  • •Add context keywords based on topic type
  • •Be specific but not overly narrow
  • •Examples:
    • •Good: "TypeScript 최신 트렌드 2024 2025"
    • •Good: "climate change latest research 2024"
    • •Good: "人工智能发展趋势 2024" (Chinese example)
    • •Too narrow: "TypeScript 5.3.2 bug fixes"
    • •Too broad: "technology"

Available parameters:

python
await web_search({
    "query": "search terms here",
    "max_results": 10,  # default: 10
    "engine": "duckduckgo"  # default, can also use "brave" or "serper"
})

Phase 2: Deep Source Analysis

Fetch and analyze 2-4 most relevant sources for detailed information.

Steps:

  1. •

    Select sources based on:

    • •Recency (prefer recent content when researching trends)
    • •Authority (official sources, major publications, recognized experts)
    • •Comprehensiveness (detailed content over brief mentions)
    • •Relevance to user's specific question
    • •Language match (prefer sources in user's language when available)
  2. •

    Use fetch tool to retrieve full content:

    • •Set appropriate max_length (6000-10000 for detailed articles)
    • •Fetch multiple sources when possible
    • •Handle fetch failures gracefully (some sites may block access)
    • •Prioritize sources that provide different perspectives
  3. •

    Extract key information:

    • •Main points and themes
    • •Specific data, statistics, or metrics
    • •Expert opinions or authoritative statements
    • •Examples or case studies
    • •Future predictions or trends

Available parameters:

python
await fetch({
    "url": "https://example.com/article",
    "max_length": 8000,  # default: 5000
    "start_index": 0,     # default: 0
    "raw": False          # default: False (returns markdown)
})

Handling fetch failures:

  • •If a source fails to fetch, continue with other sources
  • •Log the failure but don't let it stop the workflow
  • •Ensure at least 2 sources are successfully fetched
  • •If all fetches fail, synthesize from search results only

Phase 3: Information Synthesis

Organize and synthesize information into a structured document.

Language Matching:

  • •CRITICAL: Match the document language to the user's request language
  • •If user asks in Korean, write the entire document in Korean
  • •If user asks in English, write in English
  • •If user asks in another language, write in that language
  • •Maintain consistency throughout the document

Structure Guidelines:

Core sections to include:

  1. •Header: Title, date, topic overview
  2. •Executive Summary: 2-3 paragraph overview in user's language
  3. •Main Content: Organized sections with clear headings
    • •Use numbered sections or topic-based organization
    • •Include subheadings for clarity
  4. •Key Points: Important findings, data, or insights
  5. •Context: Background information, current state, significance
  6. •Conclusion: Key takeaways and synthesis
  7. •References: Source links and citations

Formatting Standards:

  • •Use clear hierarchy: Headers with ===, subheaders with ---, bullet points
  • •Include specific data: percentages, dates, numbers, names
  • •Provide examples when relevant
  • •Add visual separation: line breaks, section dividers
  • •Use symbols for emphasis: ✓, •, →, ★, 📌
  • •Keep paragraphs concise: 3-5 sentences maximum
  • •Use lists and bullet points for readability

Content Quality Standards:

  • •Accuracy: Verify claims across multiple sources
  • •Completeness: Cover all major themes identified in Phase 1
  • •Balance: Include multiple perspectives when available
  • •Specificity: Avoid vague statements; provide concrete details
  • •Clarity: Write in clear, accessible language
  • •Attribution: Cite sources for key claims and data

Document Length:

  • •Aim for comprehensive but readable documents
  • •Typical range: 150-250 lines for thorough research
  • •Adjust based on topic complexity and user needs
  • •Better to be thorough than superficial

Phase 4: File Documentation

Save the synthesized document to the user's specified file path.

Steps:

  1. •Confirm file path from user request
    • •Use absolute paths when provided
    • •Default to /tmp/ if no path specified and ask user for confirmation
  2. •Use write_file tool with mode='rewrite'
  3. •Include the complete structured document from Phase 3
  4. •Confirm successful save to user

Available parameters:

python
await write_file({
    "path": "/absolute/path/to/file.txt",
    "content": "Document content here...",
    "mode": "rewrite"  # or "append"
})

File Naming Conventions:

  • •Use descriptive names based on topic
  • •Include date if relevant: topic_2024.txt
  • •Use underscores for spaces: climate_research.txt
  • •Prefer .txt or .md extensions for text documents
  • •Match user's specified filename if provided

Best Practices

Search Strategy

  • •Cast wide net first: Initial search should be broad to capture full picture
  • •Multiple angles: Consider different search terms for comprehensive coverage
  • •Language awareness: Search in the user's language when possible
  • •Validate recency: Check publication dates when researching current topics

Source Selection

  • •Quality over quantity: Better to have 2-3 excellent sources than 5 mediocre ones
  • •Diversity: Seek different perspectives and types of sources
  • •Authority: Prioritize official sources, experts, and reputable publications
  • •Accessibility: Some sources may block fetch; have backup options

Synthesis Quality

  • •Structure matters: Well-organized content is more valuable than comprehensive chaos
  • •Context is key: Explain significance and implications, not just facts
  • •Be objective: Present balanced view and acknowledge limitations
  • •Stay focused: Address user's question directly; avoid tangents

Language and Localization

  • •Match user's language: If user asks in Korean, respond in Korean
  • •Maintain consistency: Don't mix languages within the document
  • •Cultural context: Consider cultural relevance when selecting sources
  • •Clear translation: When sources are in different languages, integrate smoothly

Common Pitfalls to Avoid

  • •Stopping too early: Don't rely only on initial search results
  • •Language mismatch: Don't respond in English when user asks in another language
  • •Poor structure: Avoid walls of text; use clear organization
  • •Missing attribution: Always cite sources for key information
  • •Ignoring failures: Handle fetch failures gracefully and continue
  • •Vague content: Provide specific information and concrete details

Quality Checklist

Before saving the file, verify:

  • •✓ Document language matches user's request language
  • •✓ All major themes from initial search are covered
  • •✓ At least 2 authoritative sources were analyzed
  • •✓ Specific data points and examples are included
  • •✓ Document follows clear structural format
  • •✓ Content is well-formatted with headers and organization
  • •✓ Conclusion provides meaningful synthesis
  • •✓ Sources are cited in references section
  • •✓ File path is correct (absolute path)
  • •✓ No placeholder text or incomplete sections

Workflow Example

User Request: "Python 머신러닝 라이브러리 트렌드를 조사해서 ~/Documents/ml_trends.txt로 저장해줘"

Execution:

  1. •

    Phase 1 - Search:

    • •Query: "Python 머신러닝 라이브러리 최신 트렌드 2024"
    • •Results: 10 sources about ML libraries, frameworks, trends
  2. •

    Phase 2 - Fetch:

    • •Fetch 2-3 Korean or English sources with detailed ML library information
    • •Extract: Library names, usage statistics, new features, comparison data
  3. •

    Phase 3 - Synthesize:

    • •Write document in Korean (matching user's language)
    • •Structure: 개요 → 주요 라이브러리 트렌드 → 비교 분석 → 결론
    • •Include: TensorFlow, PyTorch, scikit-learn statistics and trends
    • •Format: Clear headers, bullet points, data points
  4. •

    Phase 4 - Save:

    • •Write to: /Users/user/Documents/ml_trends.txt
    • •Confirm: "✅ 머신러닝 라이브러리 트렌드 리포트가 저장되었습니다!"

Cross-Domain Adaptability

This skill works across all domains and topics:

Technology: Programming languages, frameworks, tools, trends Science: Research findings, discoveries, methodologies Business: Market analysis, industry trends, company research Health: Medical research, treatments, health trends Culture: Arts, entertainment, social trends Education: Learning resources, educational trends, courses Current Events: News, developments, ongoing situations Any topic: The workflow adapts to any research subject

Multi-Language Support

This skill supports research in any language:

  • •Korean: Full support for Korean queries and Korean documents
  • •English: Full support for English queries and English documents
  • •Other languages: Adaptable to any language (Chinese, Japanese, Spanish, etc.)
  • •Mixed sources: Can synthesize from sources in multiple languages
  • •Language matching: Always match output language to user's request language

Notes

  • •Objectivity: Present balanced, factual information
  • •Timeliness: Check publication dates for time-sensitive topics
  • •User's language: Always match the document language to user's request
  • •File permissions: Verify user has write access to specified path
  • •Follow-up: Be prepared to expand or clarify sections if user requests
  • •Flexibility: Adapt structure and depth based on topic and user needs