\r \r

Deep Research Framework\r

\r A structured deep-research skill for producing rigorous, balanced, and actionable research reports on any topic.\r \r

Overview\r

\r This skill enforces a disciplined research methodology. It is NOT a simple "search and summarize" tool.\r It demands evidence quality control, counterevidence inclusion, and honest flagging of gaps.\r \r Outputs:\r

1. Structured Research Report — full-depth analysis organized by dimensions\r
2. One-Page Summary Card — key findings, confidence levels, and action items\r \r

Phase 0 — Pre-Research Clarification\r

\r Before starting research, confirm these inputs from the user or infer them from context:\r \r | Input | Required | Default if absent |\r |-------|----------|-------------------|\r | Research topic | Yes | — |\r | Target audience | Recommended | "decision maker / general professional" |\r | Key dimensions to focus on | Optional | Use standard dimension set (see references) |\r | Time scope | Optional | "recent 1–3 years unless specified" |\r | Depth level | Optional | "standard" (balanced breadth and depth) |\r \r If the topic is ambiguous, ask ONE clarifying question before proceeding.\r \r

Phase 1 — Research Design\r

\r

1.1 Define the Research Objective\r

\r State in 1–2 sentences: What question is this research answering, and what decision or understanding will it enable?\r \r

1.2 Define the Audience\r

\r State who will read this: their background, what they already know, and what they need to walk away with.\r \r

1.3 Scope and Constraints\r

\r

• Time scope: Which period does the research cover?\r
• Geographic scope: Global? China? Specific region?\r
• Exclusions: What is explicitly out of scope?\r \r

1.4 Decompose into Research Dimensions\r

\r Break the topic into 4–7 dimensions. For each dimension:\r

• Write 2–4 key questions that need answering\r
• Assign a priority: Core / Supporting / Context\r \r Use references/dimension-templates.md for standard dimension sets by research type (industry, technology, company, product, market).\r \r

Phase 2 — Multi-Source Research Execution\r

\r

2.1 Source Strategy\r

\r For each dimension, gather from at least 2–3 distinct source types:\r \r | Source Type | Examples | Default Credibility |\r |-------------|----------|---------------------|\r | Primary data | Official reports, regulatory filings, academic papers | ★★★★★ |\r | Industry reports | McKinsey, Gartner, IDC, CB Insights, Frost & Sullivan | ★★★★ |\r | News & media | Reuters, Bloomberg, FT, 36kr, 财新, 虎嗅 | ★★★ |\r | Expert commentary | Interviews, conference talks, analyst notes | ★★★ |\r | Community & forums | Reddit, Zhihu, GitHub Issues, industry forums | ★★ |\r | AI-generated synthesis | LLM summaries without cited sources | ★ |\r \r See references/credibility-rubric.md for detailed credibility rating rules.\r \r

2.2 Search Execution\r

\r For each research dimension:\r

1. Formulate targeted search queries (try multiple angles)\r
2. Retrieve results using web_search or available tools\r
3. For each piece of evidence, record: claim, source, credibility rating, date\r \r

2.3 Counterevidence Requirement (MANDATORY)\r

\r For every major claim in the report, actively search for contradicting evidence:\r

• If counterevidence exists: include it alongside the main claim, labeled [反面证据]\r
• If no counterevidence found: note [未发现明显反面证据，可信度待验]\r
• NEVER omit counterevidence to make the report cleaner or more conclusive\r \r

2.4 Conflicting Information Handling\r

\r When two credible sources contradict each other:\r

• Present BOTH positions side-by-side\r
• Note the source, date, and credibility of each\r
• Do NOT pick a side unless evidence weight is clearly asymmetric\r
• Label the conflict: [矛盾信息 — 并列呈现]\r \r

2.5 Information Gap Flagging (MANDATORY)\r

\r When a key question cannot be answered from available sources:\r

• Mark it clearly: [信息不足 — 无法得出可靠结论]\r
• State what type of source would be needed to fill the gap\r
• NEVER fabricate data, statistics, or citations to fill gaps\r \r

Phase 3 — Synthesis and Analysis\r

\r

3.1 Evidence Weighting\r

\r Synthesize findings dimension by dimension:\r

• Higher-credibility sources take precedence\r
• Recent data (within 12 months) takes precedence over older data\r
• Consensus across multiple independent sources strengthens confidence\r \r

3.2 Insight Generation\r

\r Go beyond summarizing to identify:\r

• Key patterns and trends\r
• Surprising or counterintuitive findings\r
• Implications for the research objective\r
• Open questions that remain after research\r \r

3.3 Confidence Levels\r

\r Assign a confidence level to each major finding:\r \r | Level | Meaning |\r |-------|---------|\r | 高 (High) | Multiple independent ★★★★+ sources agree |\r | 中 (Medium) | Some evidence, limited cross-validation |\r | 低 (Low) | Single source, low credibility, or conflicting signals |\r | 未知 (Unknown) | Insufficient data, explicitly flagged |\r \r

Phase 4 — Output Delivery\r

\r

4.1 Full Research Report\r

\r Follow the template in assets/report-template.md. Structure:\r \r

1. 研究概览 — objective, audience, scope, date\r
2. 执行摘要 — 5–7 bullet key findings with confidence levels\r
3. 维度分析 — one section per dimension with:\r
   • Key findings\r
   • Supporting evidence (with source + credibility)\r
   • Counterevidence\r
   • Confidence level\r
   • Information gaps (if any)\r
4. 矛盾与争议 — consolidated list of conflicting information\r
5. 综合洞察 — cross-dimension synthesis and implications\r
6. 信息缺口清单 — all flagged gaps in one place\r
7. 参考资料 — all sources with credibility ratings\r \r

4.2 One-Page Summary Card\r

\r Follow the template in assets/summary-card-template.md. Deliver after the full report.\r \r The card contains:\r

• Topic + audience + date (header)\r
• 3 most important findings (with confidence)\r
• 1 most important counterpoint\r
• 1 key uncertainty / gap\r
• Suggested next actions (if applicable)\r
• Overall research confidence score (High / Medium / Low)\r \r

Quality Checklist (self-check before delivery)\r

\r Before presenting the output, verify:\r \r

• Every major claim has a cited source with credibility rating\r
• Counterevidence section is non-empty or explicitly noted\r
• All conflicting information is presented side-by-side, not resolved by choosing one\r
• All information gaps are explicitly labeled\r
• No statistics or citations were fabricated\r
• Confidence levels are assigned to all major findings\r
• One-page summary card is included\r
• Report follows the template structure\r \r

Important Rules\r

\r

1. No fabrication: If data doesn't exist, say so. Never invent statistics, quotes, or citations.\r
2. No false certainty: Do not present uncertain findings as definitive.\r
3. Counterevidence is mandatory: A report without counterevidence is incomplete.\r
4. Conflicting sources must coexist: Do not silently pick one side.\r
5. Credibility must be visible: Every claim should be traceable to a rated source.\r