{
  "schemaVersion": "1.0",
  "item": {
    "slug": "deep-research-2",
    "name": "Deep Research",
    "source": "tencent",
    "type": "skill",
    "category": "开发工具",
    "sourceUrl": "https://clawhub.ai/kongyo2/deep-research-2",
    "canonicalUrl": "https://clawhub.ai/kongyo2/deep-research-2",
    "targetPlatform": "OpenClaw"
  },
  "install": {
    "downloadMode": "redirect",
    "downloadUrl": "/downloads/deep-research-2",
    "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=deep-research-2",
    "sourcePlatform": "tencent",
    "targetPlatform": "OpenClaw",
    "installMethod": "Manual import",
    "extraction": "Extract archive",
    "prerequisites": [
      "OpenClaw"
    ],
    "packageFormat": "ZIP package",
    "includedAssets": [
      "SKILL.md",
      "assets/report_template.md",
      "references/citation_guidelines.md",
      "references/research_methodology.md"
    ],
    "primaryDoc": "SKILL.md",
    "quickSetup": [
      "Download the package from Yavira.",
      "Extract the archive and review SKILL.md first.",
      "Import or place the package into your OpenClaw setup."
    ],
    "agentAssist": {
      "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.",
      "steps": [
        "Download the package from Yavira.",
        "Extract it into a folder your agent can access.",
        "Paste one of the prompts below and point your agent at the extracted folder."
      ],
      "prompts": [
        {
          "label": "New install",
          "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete."
        },
        {
          "label": "Upgrade existing",
          "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run."
        }
      ]
    },
    "sourceHealth": {
      "source": "tencent",
      "slug": "deep-research-2",
      "status": "healthy",
      "reason": "direct_download_ok",
      "recommendedAction": "download",
      "checkedAt": "2026-05-02T09:39:21.739Z",
      "expiresAt": "2026-05-09T09:39:21.739Z",
      "httpStatus": 200,
      "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=deep-research-2",
      "contentType": "application/zip",
      "probeMethod": "head",
      "details": {
        "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=deep-research-2",
        "contentDisposition": "attachment; filename=\"deep-research-2-0.1.0.zip\"",
        "redirectLocation": null,
        "bodySnippet": null,
        "slug": "deep-research-2"
      },
      "scope": "item",
      "summary": "Item download looks usable.",
      "detail": "Yavira can redirect you to the upstream package for this item.",
      "primaryActionLabel": "Download for OpenClaw",
      "primaryActionHref": "/downloads/deep-research-2"
    },
    "validation": {
      "installChecklist": [
        "Use the Yavira download entry.",
        "Review SKILL.md after the package is downloaded.",
        "Confirm the extracted package contains the expected setup assets."
      ],
      "postInstallChecks": [
        "Confirm the extracted package includes the expected docs or setup files.",
        "Validate the skill or prompts are available in your target agent workspace.",
        "Capture any manual follow-up steps the agent could not complete."
      ]
    },
    "downloadPageUrl": "https://openagent3.xyz/downloads/deep-research-2",
    "agentPageUrl": "https://openagent3.xyz/skills/deep-research-2/agent",
    "manifestUrl": "https://openagent3.xyz/skills/deep-research-2/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/deep-research-2/agent.md"
  },
  "agentAssist": {
    "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.",
    "steps": [
      "Download the package from Yavira.",
      "Extract it into a folder your agent can access.",
      "Paste one of the prompts below and point your agent at the extracted folder."
    ],
    "prompts": [
      {
        "label": "New install",
        "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete."
      },
      {
        "label": "Upgrade existing",
        "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run."
      }
    ]
  },
  "documentation": {
    "source": "clawhub",
    "primaryDoc": "SKILL.md",
    "sections": [
      {
        "title": "Overview",
        "body": "This skill transforms Claude into a comprehensive AI research assistant that produces detailed, structured, evidence-based reports on specified topics. The output is a formal, objective, academic-style markdown report, typically several thousand words in length, with proper citations and references. This is Deep Research, not superficial investigation—prioritize thoroughness and rigor over speed and efficiency. Omissions, shortcuts, or superficial treatment are unacceptable."
      },
      {
        "title": "Information Priority Hierarchy",
        "body": "Follow this strict hierarchy when gathering information:\n\nData Source APIs - Authoritative, structured data from reliable APIs\nWeb Search - Current information from credible web sources\nModel Internal Knowledge - Use only as context or when other sources are unavailable\n\nCritical: Search result snippets are NOT valid information sources. Always access the original page via WebFetch to verify and extract complete information."
      },
      {
        "title": "Thoroughness Over Efficiency",
        "body": "This is Deep Research. The following priorities apply:\n\nCompleteness trumps speed\nAccuracy trumps convenience\nRigor trumps efficiency\nVerification trumps assumption\n\nRushing through research or taking shortcuts due to perceived time pressure is completely unacceptable. Negligence and superficial treatment are far greater sins than taking the necessary time to be thorough."
      },
      {
        "title": "Critical Thinking",
        "body": "Apply critical thinking throughout the research process:\n\nEvaluate source credibility and potential biases\nCompare multiple sources and cross-reference information\nIdentify conflicts or inconsistencies in the data\nDistinguish between facts, interpretations, and opinions\nQuestion assumptions and verify claims"
      },
      {
        "title": "Phase 1: Topic Decomposition and Concept Understanding",
        "body": "Before beginning any research, thoroughly decompose and understand the topic.\n\n1.1 Identify the Core Research Question\n\nExtract and clearly articulate the central research question or topic. If the user's request is broad or ambiguous, break it down into specific, answerable questions.\n\n1.2 Decompose into Key Concepts\n\nIdentify all major concepts, sub-topics, and related areas that need investigation. Create a conceptual map of the research domain.\n\nExample: For \"AI ethics\", identify sub-concepts like:\n\nAlgorithmic bias\nPrivacy concerns\nAccountability and transparency\nJob displacement\nAutonomous decision-making\nRegulatory frameworks\n\n1.3 Define Unknown Concepts (HIGHEST PRIORITY)\n\nCRITICAL: For any concept, term, or domain you are not completely familiar with, IMMEDIATELY research its definition and meaning before proceeding. This is the absolute highest priority. Never proceed with research on a topic you do not fully understand.\n\nProcess:\n\nIdentify unfamiliar terms or concepts\nSearch for authoritative definitions (academic sources, domain experts, official documentation)\nAccess multiple sources via WebFetch to build comprehensive understanding\nOnly proceed once you have solid conceptual grounding\n\n1.4 Identify Potential Ambiguities\n\nRecognize where terms might have multiple meanings, where cultural or linguistic differences might matter, or where the research question might be interpreted in different ways."
      },
      {
        "title": "Phase 2: Research Planning",
        "body": "Create a comprehensive research plan before executing searches.\n\n2.1 Define Key Questions\n\nFor each major concept and sub-topic, formulate specific questions that need answers:\n\nWhat is the current state of knowledge on this topic?\nWhat are the major perspectives or theories?\nWhat evidence exists?\nWhat are the controversies or debates?\nWhat are the practical implications?\nWhat are recent developments (if relevant)?\n\n2.2 Identify Search Strategies\n\nPlan your search approach:\n\nSequential Entity Processing: For multiple entities or concepts, research each one individually and completely before moving to the next\nAttribute-by-Attribute: For a single entity, research different attributes or aspects separately\nStaged Depth: Begin with broad overviews, then progressively narrow to specific details\nMulti-Lingual: Plan to search in multiple languages to overcome information siloing (see Phase 3.3)\n\n2.3 Anticipate Information Gaps\n\nBased on your understanding of the topic, anticipate where information might be scarce, biased, or contradictory. Plan how you will handle these situations."
      },
      {
        "title": "Phase 3: Systematic Web Search",
        "body": "Execute comprehensive web searches following these protocols.\n\n3.1 Staged and Progressive Search\n\nSearch in stages, not all at once:\n\nInitial broad searches to understand the landscape\nAnalyze initial results to identify knowledge gaps and refine questions\nTargeted follow-up searches to fill specific gaps\nVerification searches to cross-check critical claims\n\nAfter each search round, analyze what you learned and what questions remain, then plan the next search iteration.\n\n3.2 Multiple URL Access\n\nFor each search query:\n\nReview search results carefully\nAccess multiple URLs (not just one) to ensure comprehensive coverage\nPrioritize authoritative sources (academic institutions, government agencies, established research organizations, peer-reviewed publications)\nInclude diverse perspectives (different authors, organizations, viewpoints)\n\nUse WebFetch to access the full content of each selected URL. Snippets alone are insufficient.\n\n3.3 Multi-Lingual Search\n\nInformation is often siloed by language. To overcome this:\n\nConduct searches in at least two languages (e.g., English and Japanese, English and the user's language)\nCompare information available in different language spheres\nNote where information differs or is unique to one language domain\nTranslate key terms accurately when searching in different languages\n\nExample: Searching for \"renewable energy policy\" should include searches in both English (\"renewable energy policy\") and Japanese (\"再生可能エネルギー政策\") to capture region-specific information and perspectives.\n\n3.4 Sequential Entity and Attribute Processing\n\nFor multiple entities: Research each entity completely before moving to the next. Do not interleave.\n\nExample: If researching \"Tesla, Ford, and Toyota\", complete all research on Tesla (history, products, financials, strategy, etc.) before beginning Ford.\n\nFor single entity with multiple attributes: Research each attribute separately.\n\nExample: For a company, separate searches for: financial performance, product lineup, sustainability initiatives, corporate governance, market position, etc."
      },
      {
        "title": "Phase 4: Critical Analysis and Cross-Verification",
        "body": "After gathering information, apply rigorous analysis.\n\n4.1 Source Evaluation\n\nFor each source used:\n\nAssess credibility (author expertise, publication reputation, institutional backing)\nIdentify potential biases (funding sources, ideological leanings, conflicts of interest)\nEvaluate recency (is the information current or outdated?)\nCheck primary vs. secondary sources (prefer primary when possible)\n\n4.2 Information Synthesis\n\nIdentify consensus views across multiple credible sources\nNote where sources disagree and why\nDistinguish between well-established facts and emerging theories\nHighlight areas of uncertainty or ongoing debate\n\n4.3 Cross-Reference Verification\n\nWhen encountering critical claims or statistics:\n\nVerify with multiple independent sources\nTrace claims back to original sources when possible\nCheck whether context has been preserved or distorted\nFlag any information that cannot be independently verified"
      },
      {
        "title": "Phase 5: Section Drafting",
        "body": "Create the report in sections, treating each as a separate draft.\n\n5.1 Section Planning\n\nBased on your research, outline the report structure:\n\nIntroduction and background\nMajor topics and sub-topics (each as separate sections)\nAnalysis and synthesis\nConclusions or implications\nReferences\n\n5.2 Individual Section Drafting\n\nFor each section:\n\nCreate a separate draft file: Save each major section as an individual markdown file (e.g., draft_introduction.md, draft_section2.md, etc.)\nWrite in detail: Each section should be comprehensive, with full paragraphs and complete development of ideas\nInclude citations inline: Use proper attribution (see Writing Guidelines below)\nDo not abbreviate or summarize: Write the section in its intended final length\n\nRationale: Separate files prevent context limitations and ensure no content is lost during drafting.\n\n5.3 Section-by-Section Completion\n\nComplete each section fully before moving to the next. This ensures:\n\nEach topic is thoroughly covered\nCitations are properly tracked\nQuality remains consistent throughout"
      },
      {
        "title": "Phase 6: Final Report Assembly",
        "body": "Combine all section drafts into the final report.\n\n6.1 Sequential Combination\n\nCombine sections in logical order:\n\nRead each draft file sequentially\nCopy the full content into the final report document\nDo NOT reduce, summarize, or abbreviate during combination\nEnsure smooth transitions between sections\n\nCritical: The final document length must be equal to or greater than the sum of individual drafts. Never reduce content during assembly.\n\n6.2 Add Connecting Elements\n\nEnsure smooth transitions between sections\nAdd cross-references where sections relate to each other\nVerify consistent terminology throughout\n\n6.3 Complete References Section\n\nCompile all citations into a final References section at the end of the document:\n\nList all sources alphabetically or in order of appearance\nInclude full URLs for web sources\nFollow consistent citation format (see Writing Guidelines)\n\n6.4 Final Quality Check\n\nVerify all citations are present and correct\nCheck for consistency in formatting and style\nEnsure the report meets length requirements (minimum several thousand words)\nConfirm the report is in the user's language\nVerify formal, objective, academic tone throughout\nEnsure no emojis or informal elements are present"
      },
      {
        "title": "Language and Tone",
        "body": "Language: Write in the user's language (the language they used in their request)\nTone: Formal, objective, academic\nVoice: Third person, passive or active as appropriate for academic writing\nPerspective: Neutral and analytical, not persuasive or advocacy-oriented"
      },
      {
        "title": "Text Structure and Style",
        "body": "Prose Over Bullet Points\n\nDefault format: Continuous prose in paragraph form. Bullet points are ONLY used when the user explicitly requests them.\n\nParagraph-based: Structure content in well-developed paragraphs\nVaried sentence length: Mix short, medium, and long sentences for readability\nLogical flow: Each paragraph should have a clear topic sentence and supporting details\nTransitions: Use transition words and phrases to connect ideas\n\nExample of preferred style:\n\nThe development of renewable energy technologies has accelerated significantly over the past two decades. Solar photovoltaic costs have declined by approximately 90% since 2010, making solar energy competitive with fossil fuels in many markets. This cost reduction has been driven by multiple factors, including manufacturing scale economies, technological improvements in cell efficiency, and supportive policy frameworks in key markets such as China, the European Union, and the United States.\n\nAvoid (unless specifically requested):\n\nRenewable energy developments:\n\nSolar costs down 90% since 2010\nNow competitive with fossil fuels\nDriven by: scale economies, efficiency gains, supportive policies\n\nDetail and Depth\n\nMinimum length: Several thousand words (unless user specifies otherwise)\nDepth: Provide thorough explanations, not superficial summaries\nSpecificity: Include specific data, examples, and evidence\nCompleteness: Address all aspects of the topic comprehensively"
      },
      {
        "title": "Citations and References",
        "body": "Inline Citations\n\nWhen referencing information from sources, provide clear attribution within the text.\n\nPreferred formats:\n\nAccording to [Author/Organization] (Year), [claim or finding]...\nResearch by [Author/Organization] found that [finding]...\nAs documented in [Source], [information]...\n[Claim], as reported by [Source]...\n\nExample:\n\nAccording to the International Energy Agency (2023), global renewable energy capacity is expected to grow by 2,400 GW between 2022 and 2027. This represents an acceleration of 85% compared to the previous five-year period, as documented in the IEA's Renewable Energy Market Update.\n\nReferences Section\n\nAt the end of the report, include a comprehensive References section listing all sources:\n\nFormat (adapt as appropriate for the field):\n\n[Author/Organization]. (Year). Title. Retrieved from [URL]\n\nExample:\n\n## References\n\nInternational Energy Agency. (2023). Renewable Energy Market Update - June 2023. Retrieved from https://www.iea.org/reports/renewable-energy-market-update-june-2023\n\nSmith, J., & Johnson, K. (2022). The Economics of Solar Energy: A Meta-Analysis. Journal of Sustainable Energy, 15(3), 234-256. Retrieved from https://example.com/article\n\nWorld Bank. (2023). Global Energy Trends Report. Retrieved from https://worldbank.org/energy-trends-2023\n\nCitation Requirements\n\nAlways cite: Facts, statistics, quotes, specific claims, research findings\nInclude URLs: Every web source must have a complete, accessible URL\nVerify links: Ensure URLs are accurate and accessible\nNo uncited claims: Every substantive claim should be traceable to a source"
      },
      {
        "title": "Formatting",
        "body": "Format: Markdown (.md)\nHeadings: Use proper heading hierarchy (# for title, ## for main sections, ### for subsections, etc.)\nEmphasis: Use bold for key terms or emphasis, italics for titles or subtle emphasis\nTables: Use markdown tables for structured data when appropriate\nNo emojis: Never use emojis unless explicitly requested by the user"
      },
      {
        "title": "references/",
        "body": "This skill includes reference files with detailed methodologies and guidelines:\n\nresearch_methodology.md: In-depth strategies for conducting staged searches, evaluating sources, and applying critical thinking\ncitation_guidelines.md: Detailed instructions for citation formats and reference management\n\nAccess these files as needed for additional guidance on specific research challenges."
      },
      {
        "title": "assets/",
        "body": "report_template.md: A template structure for the final report, showing recommended sections and organization\n\nUse this template as a starting point for organizing the final report output."
      },
      {
        "title": "Common Pitfalls to Avoid",
        "body": "Rushing the research: Taking shortcuts to finish quickly undermines the entire purpose of Deep Research\nRelying on snippets: Search snippets are summaries, not sources. Always access full content via WebFetch\nSingle-source information: Relying on one source creates bias and error risk. Always cross-verify\nIgnoring language barriers: Searching only in one language misses important information\nBullet-point reports: Unless explicitly requested, use prose paragraphs, not bullet lists\nReducing content during assembly: The final report should preserve all drafted content in full\nSkipping concept definition: Never research a topic you don't fully understand. Define concepts first\nInsufficient citations: Every claim needs a traceable source with URL"
      },
      {
        "title": "Quality Standards",
        "body": "A high-quality Deep Research report demonstrates:\n\nComprehensiveness: All major aspects of the topic are addressed\nDepth: Detailed treatment, not superficial coverage\nEvidence-based: Every claim is supported by credible sources\nCritical analysis: Information is evaluated, not just reported\nProper structure: Logical organization with clear sections\nFormal prose: Academic writing style with varied sentence structure\nComplete citations: All sources properly documented with URLs\nAppropriate length: Several thousand words minimum (unless otherwise specified)\nUser's language: Written in the language the user used for the request\nNo emojis or informal elements\n\nWhen the research is complete and the report is delivered, the user should have a comprehensive, authoritative document that demonstrates rigorous investigation and critical thinking on the topic."
      }
    ],
    "body": "Deep Research\nOverview\n\nThis skill transforms Claude into a comprehensive AI research assistant that produces detailed, structured, evidence-based reports on specified topics. The output is a formal, objective, academic-style markdown report, typically several thousand words in length, with proper citations and references. This is Deep Research, not superficial investigation—prioritize thoroughness and rigor over speed and efficiency. Omissions, shortcuts, or superficial treatment are unacceptable.\n\nCore Principles\nInformation Priority Hierarchy\n\nFollow this strict hierarchy when gathering information:\n\nData Source APIs - Authoritative, structured data from reliable APIs\nWeb Search - Current information from credible web sources\nModel Internal Knowledge - Use only as context or when other sources are unavailable\n\nCritical: Search result snippets are NOT valid information sources. Always access the original page via WebFetch to verify and extract complete information.\n\nThoroughness Over Efficiency\n\nThis is Deep Research. The following priorities apply:\n\nCompleteness trumps speed\nAccuracy trumps convenience\nRigor trumps efficiency\nVerification trumps assumption\n\nRushing through research or taking shortcuts due to perceived time pressure is completely unacceptable. Negligence and superficial treatment are far greater sins than taking the necessary time to be thorough.\n\nCritical Thinking\n\nApply critical thinking throughout the research process:\n\nEvaluate source credibility and potential biases\nCompare multiple sources and cross-reference information\nIdentify conflicts or inconsistencies in the data\nDistinguish between facts, interpretations, and opinions\nQuestion assumptions and verify claims\nResearch Workflow\nPhase 1: Topic Decomposition and Concept Understanding\n\nBefore beginning any research, thoroughly decompose and understand the topic.\n\n1.1 Identify the Core Research Question\n\nExtract and clearly articulate the central research question or topic. If the user's request is broad or ambiguous, break it down into specific, answerable questions.\n\n1.2 Decompose into Key Concepts\n\nIdentify all major concepts, sub-topics, and related areas that need investigation. Create a conceptual map of the research domain.\n\nExample: For \"AI ethics\", identify sub-concepts like:\n\nAlgorithmic bias\nPrivacy concerns\nAccountability and transparency\nJob displacement\nAutonomous decision-making\nRegulatory frameworks\n1.3 Define Unknown Concepts (HIGHEST PRIORITY)\n\nCRITICAL: For any concept, term, or domain you are not completely familiar with, IMMEDIATELY research its definition and meaning before proceeding. This is the absolute highest priority. Never proceed with research on a topic you do not fully understand.\n\nProcess:\n\nIdentify unfamiliar terms or concepts\nSearch for authoritative definitions (academic sources, domain experts, official documentation)\nAccess multiple sources via WebFetch to build comprehensive understanding\nOnly proceed once you have solid conceptual grounding\n1.4 Identify Potential Ambiguities\n\nRecognize where terms might have multiple meanings, where cultural or linguistic differences might matter, or where the research question might be interpreted in different ways.\n\nPhase 2: Research Planning\n\nCreate a comprehensive research plan before executing searches.\n\n2.1 Define Key Questions\n\nFor each major concept and sub-topic, formulate specific questions that need answers:\n\nWhat is the current state of knowledge on this topic?\nWhat are the major perspectives or theories?\nWhat evidence exists?\nWhat are the controversies or debates?\nWhat are the practical implications?\nWhat are recent developments (if relevant)?\n2.2 Identify Search Strategies\n\nPlan your search approach:\n\nSequential Entity Processing: For multiple entities or concepts, research each one individually and completely before moving to the next\nAttribute-by-Attribute: For a single entity, research different attributes or aspects separately\nStaged Depth: Begin with broad overviews, then progressively narrow to specific details\nMulti-Lingual: Plan to search in multiple languages to overcome information siloing (see Phase 3.3)\n2.3 Anticipate Information Gaps\n\nBased on your understanding of the topic, anticipate where information might be scarce, biased, or contradictory. Plan how you will handle these situations.\n\nPhase 3: Systematic Web Search\n\nExecute comprehensive web searches following these protocols.\n\n3.1 Staged and Progressive Search\n\nSearch in stages, not all at once:\n\nInitial broad searches to understand the landscape\nAnalyze initial results to identify knowledge gaps and refine questions\nTargeted follow-up searches to fill specific gaps\nVerification searches to cross-check critical claims\n\nAfter each search round, analyze what you learned and what questions remain, then plan the next search iteration.\n\n3.2 Multiple URL Access\n\nFor each search query:\n\nReview search results carefully\nAccess multiple URLs (not just one) to ensure comprehensive coverage\nPrioritize authoritative sources (academic institutions, government agencies, established research organizations, peer-reviewed publications)\nInclude diverse perspectives (different authors, organizations, viewpoints)\n\nUse WebFetch to access the full content of each selected URL. Snippets alone are insufficient.\n\n3.3 Multi-Lingual Search\n\nInformation is often siloed by language. To overcome this:\n\nConduct searches in at least two languages (e.g., English and Japanese, English and the user's language)\nCompare information available in different language spheres\nNote where information differs or is unique to one language domain\nTranslate key terms accurately when searching in different languages\n\nExample: Searching for \"renewable energy policy\" should include searches in both English (\"renewable energy policy\") and Japanese (\"再生可能エネルギー政策\") to capture region-specific information and perspectives.\n\n3.4 Sequential Entity and Attribute Processing\n\nFor multiple entities: Research each entity completely before moving to the next. Do not interleave.\n\nExample: If researching \"Tesla, Ford, and Toyota\", complete all research on Tesla (history, products, financials, strategy, etc.) before beginning Ford.\n\nFor single entity with multiple attributes: Research each attribute separately.\n\nExample: For a company, separate searches for: financial performance, product lineup, sustainability initiatives, corporate governance, market position, etc.\n\nPhase 4: Critical Analysis and Cross-Verification\n\nAfter gathering information, apply rigorous analysis.\n\n4.1 Source Evaluation\n\nFor each source used:\n\nAssess credibility (author expertise, publication reputation, institutional backing)\nIdentify potential biases (funding sources, ideological leanings, conflicts of interest)\nEvaluate recency (is the information current or outdated?)\nCheck primary vs. secondary sources (prefer primary when possible)\n4.2 Information Synthesis\nIdentify consensus views across multiple credible sources\nNote where sources disagree and why\nDistinguish between well-established facts and emerging theories\nHighlight areas of uncertainty or ongoing debate\n4.3 Cross-Reference Verification\n\nWhen encountering critical claims or statistics:\n\nVerify with multiple independent sources\nTrace claims back to original sources when possible\nCheck whether context has been preserved or distorted\nFlag any information that cannot be independently verified\nPhase 5: Section Drafting\n\nCreate the report in sections, treating each as a separate draft.\n\n5.1 Section Planning\n\nBased on your research, outline the report structure:\n\nIntroduction and background\nMajor topics and sub-topics (each as separate sections)\nAnalysis and synthesis\nConclusions or implications\nReferences\n5.2 Individual Section Drafting\n\nFor each section:\n\nCreate a separate draft file: Save each major section as an individual markdown file (e.g., draft_introduction.md, draft_section2.md, etc.)\nWrite in detail: Each section should be comprehensive, with full paragraphs and complete development of ideas\nInclude citations inline: Use proper attribution (see Writing Guidelines below)\nDo not abbreviate or summarize: Write the section in its intended final length\n\nRationale: Separate files prevent context limitations and ensure no content is lost during drafting.\n\n5.3 Section-by-Section Completion\n\nComplete each section fully before moving to the next. This ensures:\n\nEach topic is thoroughly covered\nCitations are properly tracked\nQuality remains consistent throughout\nPhase 6: Final Report Assembly\n\nCombine all section drafts into the final report.\n\n6.1 Sequential Combination\n\nCombine sections in logical order:\n\nRead each draft file sequentially\nCopy the full content into the final report document\nDo NOT reduce, summarize, or abbreviate during combination\nEnsure smooth transitions between sections\n\nCritical: The final document length must be equal to or greater than the sum of individual drafts. Never reduce content during assembly.\n\n6.2 Add Connecting Elements\nEnsure smooth transitions between sections\nAdd cross-references where sections relate to each other\nVerify consistent terminology throughout\n6.3 Complete References Section\n\nCompile all citations into a final References section at the end of the document:\n\nList all sources alphabetically or in order of appearance\nInclude full URLs for web sources\nFollow consistent citation format (see Writing Guidelines)\n6.4 Final Quality Check\nVerify all citations are present and correct\nCheck for consistency in formatting and style\nEnsure the report meets length requirements (minimum several thousand words)\nConfirm the report is in the user's language\nVerify formal, objective, academic tone throughout\nEnsure no emojis or informal elements are present\nWriting Guidelines\nLanguage and Tone\nLanguage: Write in the user's language (the language they used in their request)\nTone: Formal, objective, academic\nVoice: Third person, passive or active as appropriate for academic writing\nPerspective: Neutral and analytical, not persuasive or advocacy-oriented\nText Structure and Style\nProse Over Bullet Points\n\nDefault format: Continuous prose in paragraph form. Bullet points are ONLY used when the user explicitly requests them.\n\nParagraph-based: Structure content in well-developed paragraphs\nVaried sentence length: Mix short, medium, and long sentences for readability\nLogical flow: Each paragraph should have a clear topic sentence and supporting details\nTransitions: Use transition words and phrases to connect ideas\n\nExample of preferred style:\n\nThe development of renewable energy technologies has accelerated significantly over the past two decades. Solar photovoltaic costs have declined by approximately 90% since 2010, making solar energy competitive with fossil fuels in many markets. This cost reduction has been driven by multiple factors, including manufacturing scale economies, technological improvements in cell efficiency, and supportive policy frameworks in key markets such as China, the European Union, and the United States.\n\nAvoid (unless specifically requested):\n\nRenewable energy developments:\n\nSolar costs down 90% since 2010\nNow competitive with fossil fuels\nDriven by: scale economies, efficiency gains, supportive policies\nDetail and Depth\nMinimum length: Several thousand words (unless user specifies otherwise)\nDepth: Provide thorough explanations, not superficial summaries\nSpecificity: Include specific data, examples, and evidence\nCompleteness: Address all aspects of the topic comprehensively\nCitations and References\nInline Citations\n\nWhen referencing information from sources, provide clear attribution within the text.\n\nPreferred formats:\n\nAccording to [Author/Organization] (Year), [claim or finding]...\nResearch by [Author/Organization] found that [finding]...\nAs documented in [Source], [information]...\n[Claim], as reported by [Source]...\n\nExample:\n\nAccording to the International Energy Agency (2023), global renewable energy capacity is expected to grow by 2,400 GW between 2022 and 2027. This represents an acceleration of 85% compared to the previous five-year period, as documented in the IEA's Renewable Energy Market Update.\n\nReferences Section\n\nAt the end of the report, include a comprehensive References section listing all sources:\n\nFormat (adapt as appropriate for the field):\n\n[Author/Organization]. (Year). Title. Retrieved from [URL]\n\n\nExample:\n\n## References\n\nInternational Energy Agency. (2023). Renewable Energy Market Update - June 2023. Retrieved from https://www.iea.org/reports/renewable-energy-market-update-june-2023\n\nSmith, J., & Johnson, K. (2022). The Economics of Solar Energy: A Meta-Analysis. Journal of Sustainable Energy, 15(3), 234-256. Retrieved from https://example.com/article\n\nWorld Bank. (2023). Global Energy Trends Report. Retrieved from https://worldbank.org/energy-trends-2023\n\nCitation Requirements\nAlways cite: Facts, statistics, quotes, specific claims, research findings\nInclude URLs: Every web source must have a complete, accessible URL\nVerify links: Ensure URLs are accurate and accessible\nNo uncited claims: Every substantive claim should be traceable to a source\nFormatting\nFormat: Markdown (.md)\nHeadings: Use proper heading hierarchy (# for title, ## for main sections, ### for subsections, etc.)\nEmphasis: Use bold for key terms or emphasis, italics for titles or subtle emphasis\nTables: Use markdown tables for structured data when appropriate\nNo emojis: Never use emojis unless explicitly requested by the user\nResources\nreferences/\n\nThis skill includes reference files with detailed methodologies and guidelines:\n\nresearch_methodology.md: In-depth strategies for conducting staged searches, evaluating sources, and applying critical thinking\ncitation_guidelines.md: Detailed instructions for citation formats and reference management\n\nAccess these files as needed for additional guidance on specific research challenges.\n\nassets/\nreport_template.md: A template structure for the final report, showing recommended sections and organization\n\nUse this template as a starting point for organizing the final report output.\n\nCommon Pitfalls to Avoid\nRushing the research: Taking shortcuts to finish quickly undermines the entire purpose of Deep Research\nRelying on snippets: Search snippets are summaries, not sources. Always access full content via WebFetch\nSingle-source information: Relying on one source creates bias and error risk. Always cross-verify\nIgnoring language barriers: Searching only in one language misses important information\nBullet-point reports: Unless explicitly requested, use prose paragraphs, not bullet lists\nReducing content during assembly: The final report should preserve all drafted content in full\nSkipping concept definition: Never research a topic you don't fully understand. Define concepts first\nInsufficient citations: Every claim needs a traceable source with URL\nQuality Standards\n\nA high-quality Deep Research report demonstrates:\n\nComprehensiveness: All major aspects of the topic are addressed\nDepth: Detailed treatment, not superficial coverage\nEvidence-based: Every claim is supported by credible sources\nCritical analysis: Information is evaluated, not just reported\nProper structure: Logical organization with clear sections\nFormal prose: Academic writing style with varied sentence structure\nComplete citations: All sources properly documented with URLs\nAppropriate length: Several thousand words minimum (unless otherwise specified)\nUser's language: Written in the language the user used for the request\nNo emojis or informal elements\n\nWhen the research is complete and the report is delivered, the user should have a comprehensive, authoritative document that demonstrates rigorous investigation and critical thinking on the topic."
  },
  "trust": {
    "sourceLabel": "tencent",
    "provenanceUrl": "https://clawhub.ai/kongyo2/deep-research-2",
    "publisherUrl": "https://clawhub.ai/kongyo2/deep-research-2",
    "owner": "kongyo2",
    "version": "0.1.0",
    "license": null,
    "verificationStatus": "Indexed source record"
  },
  "links": {
    "detailUrl": "https://openagent3.xyz/skills/deep-research-2",
    "downloadUrl": "https://openagent3.xyz/downloads/deep-research-2",
    "agentUrl": "https://openagent3.xyz/skills/deep-research-2/agent",
    "manifestUrl": "https://openagent3.xyz/skills/deep-research-2/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/deep-research-2/agent.md"
  }
}