deep-researcher

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Deep Researcher v2.0

Deep Researcher v2.0

Comprehensive research methodology with file-based tracking, parallel execution, and context management for investigations requiring 5+ sources.
CRITICAL: All medical evidence and citations must come from PubMed MCP. No exceptions.

针对需要 5 个以上来源的调查,提供一套包含基于文件的跟踪、并行执行和上下文管理的综合研究方法。
关键:所有医学证据和引文必须来自 PubMed MCP,不设例外。

Research Modes

研究模式

Quick Research (1-4 sources): Work in-context, no file structure needed.
Deep Research (5+ sources): Use file-based tracking below.

快速研究(Quick Research)(1-4 个来源):在当前上下文中工作,无需文件结构。
深度研究(Deep Research)(5 个以上来源):使用下文的基于文件的跟踪方式。

Research Sources (STRICT POLICY)

研究来源(严格政策)

ALLOWED for Medical Citations

医学引文允许使用的来源

SourceToolUse Case
PubMed MCP
pubmed_search_articles
,
pubmed_fetch_contents
,
pubmed_article_connections
ALL medical evidence, trials, mechanisms
Official Guidelines
web_fetch
to ACC/ESC/ADA/AHA URLs only
Guideline recommendations
AstraDB RAGKnowledge pipelineTextbook references, pre-loaded guidelines
来源工具用途
PubMed MCP
pubmed_search_articles
、
pubmed_fetch_contents
、
pubmed_article_connections
所有医学证据、试验和机制
官方指南仅通过
web_fetch
访问 ACC/ESC/ADA/AHA URL
指南建议
AstraDB RAG知识流水线教科书参考、预加载指南

NOT ALLOWED for Medical Citations

医学引文不允许使用的来源

SourceWhy ExcludedAllowed Use
OpenAlexQuality variableREMOVED
PerplexityNot peer-reviewedTrend discovery only, NEVER cite
General web searchUnreliableTopic discovery only, NEVER cite
News articlesNot primary evidenceBackground context only
来源排除原因允许用途
OpenAlex质量不稳定已移除
Perplexity未经过同行评审仅用于发现趋势,绝不可引用
通用 Web 搜索可靠性不足仅用于发现主题,绝不可引用
新闻文章非一手证据仅用于背景信息

PubMed Quality Filters

PubMed 质量筛选标准

Prefer (Tier 1):
  • Randomized Controlled Trials (RCTs)
  • Meta-analyses and Systematic Reviews
  • Guidelines from ACC/ESC/ADA/AHA
Accept (Tier 2):
  • Large observational studies from Q1 journals
  • Cohort studies with >1000 patients
  • Registry data from established registries
Use Cautiously (Tier 3):
  • Case series (only if no better evidence)
  • Expert consensus statements
  • Narrative reviews (as background, not primary evidence)
Reject:
  • Case reports (except for rare conditions)
  • Letters to editor
  • Preprints without peer review
  • Animal studies (unless specifically about mechanisms)

优先采用(第 1 层):
  • 随机对照试验(RCT)
  • Meta 分析和系统综述
  • ACC/ESC/ADA/AHA 发布的指南
可接受(第 2 层):
  • 来自 Q1 期刊的大型观察性研究
  • 纳入超过 1000 名患者的队列研究
  • 成熟注册平台中的注册数据
谨慎使用(第 3 层):
  • 病例系列(仅在没有更好证据时使用)
  • 专家共识声明
  • 叙述性综述(仅作为背景信息,不作为主要证据)
不采用:
  • 病例报告(罕见疾病除外)
  • 致编辑的信件
  • 未经过同行评审的预印本
  • 动物研究(除非专门研究机制)

Deep Research Workflow

深度研究工作流

Progress Tracking

进度跟踪

Create this checklist and update after each step:
Deep Research Progress:
- [ ] Step 1: Initialize research project
- [ ] Step 2: Define scope and plan
- [ ] Step 3: Execute research threads (parallel when possible)
- [ ] Step 4: Validate and cross-reference
- [ ] Step 5: Synthesize from files
- [ ] Step 6: Generate final report

创建以下清单,并在每一步完成后更新:
Deep Research Progress:
- [ ] Step 1: Initialize research project
- [ ] Step 2: Define scope and plan
- [ ] Step 3: Execute research threads (parallel when possible)
- [ ] Step 4: Validate and cross-reference
- [ ] Step 5: Synthesize from files
- [ ] Step 6: Generate final report

Step 1: Initialize Research Project

第 1 步:初始化研究项目

For research requiring 5+ sources, create a project structure:
bash
mkdir -p ~/research_{topic}/sources
mkdir -p ~/research_{topic}/threads
Project Structure:
~/research_{topic}/
├── plan.md              # Research questions, scope, thread assignments
├── progress.md          # Living checklist, updated throughout
├── sources/
│   └── pubmed.md        # PubMed search results and abstracts
├── threads/
│   ├── thread_1.md      # Independent research thread
│   ├── thread_2.md      # Another thread
│   └── ...
├── validation.md        # Cross-reference and credibility check
├── synthesis.md         # Cross-thread analysis
└── report.md            # Final deliverable
Why file-based? Context windows fill up. Writing findings to files lets you:
  • Continue researching without context pressure
  • Synthesize from persistent storage, not memory
  • Produce larger, more comprehensive reports
  • Resume if interrupted

对于需要 5 个以上来源的研究,创建如下项目结构:
bash
mkdir -p ~/research_{topic}/sources
mkdir -p ~/research_{topic}/threads
项目结构:
~/research_{topic}/
├── plan.md              # Research questions, scope, thread assignments
├── progress.md          # Living checklist, updated throughout
├── sources/
│   └── pubmed.md        # PubMed search results and abstracts
├── threads/
│   ├── thread_1.md      # Independent research thread
│   ├── thread_2.md      # Another thread
│   └── ...
├── validation.md        # Cross-reference and credibility check
├── synthesis.md         # Cross-thread analysis
└── report.md            # Final deliverable
为什么要采用基于文件的方式? 上下文窗口会被占满。将研究发现写入文件可以让你:
  • 在不受上下文压力影响的情况下继续研究
  • 基于持久化存储而不是记忆进行综合
  • 生成规模更大、更全面的报告
  • 在工作被中断后恢复

Step 2: Define Scope and Research Plan

第 2 步:定义范围并制定研究计划

Write
plan.md
with:
markdown
undefined
在
plan.md
中编写:
markdown
undefined

Research Plan: {Topic}

Research Plan: {Topic}

Primary Question

Primary Question

[The main thing we're trying to answer]
[The main thing we're trying to answer]

Scope

Scope

  • Include: [what's in scope]
  • Exclude: [what's explicitly out]
  • Depth: [overview | detailed | exhaustive]
  • Deliverable: [report type and length]
  • Include: [what's in scope]
  • Exclude: [what's explicitly out]
  • Depth: [overview | detailed | exhaustive]
  • Deliverable: [report type and length]

Research Threads

Research Threads

Thread 1: {Subtopic A}

Thread 1: {Subtopic A}

  • Questions to answer: ...
  • PubMed search strategy: [MeSH terms, filters]
  • Expected study types: RCTs, meta-analyses, etc.
  • Can run parallel? Yes/No
  • Questions to answer: ...
  • PubMed search strategy: [MeSH terms, filters]
  • Expected study types: RCTs, meta-analyses, etc.
  • Can run parallel? Yes/No

Thread 2: {Subtopic B}

Thread 2: {Subtopic B}

  • Questions to answer: ...
  • PubMed search strategy: ...
  • Can run parallel? Yes/No
[Continue for 2-5 threads]
  • Questions to answer: ...
  • PubMed search strategy: ...
  • Can run parallel? Yes/No
[Continue for 2-5 threads]

Thread Dependencies

Thread Dependencies

  • Thread 3 depends on Thread 1 findings
  • Threads 1, 2, 4 can run in parallel
  • Thread 3 depends on Thread 1 findings
  • Threads 1, 2, 4 can run in parallel

Synthesis Strategy

Synthesis Strategy

How will threads combine into final answer?

**Planning Guidelines:**
| Research Type | Threads | Pattern |
|--------------|---------|---------|
| Simple fact-finding | 1-2 | Sequential |
| Drug comparison | 1 per drug (max 5) | Parallel |
| Complex investigation | 3-5 thematic | Mixed |
| Literature review | By time period or theme | Sequential |

---
How will threads combine into final answer?

**规划指南:**
| 研究类型 | 线程数 | 模式 |
|--------------|---------|---------|
| 简单事实查找 | 1-2 | 顺序 |
| 药物比较 | 每种药物 1 个(最多 5 个) | 并行 |
| 复杂调查 | 3-5 个主题 | 混合 |
| 文献综述 | 按时间段或主题划分 | 顺序 |

---

Step 3: Execute Research Threads

第 3 步:执行研究线程

PubMed Search Strategy

PubMed 搜索策略

For each thread, use structured PubMed queries:
python
undefined
对于每个线程,使用结构化的 PubMed 查询:
python
undefined

Example search for SGLT2 CV outcomes

Example search for SGLT2 CV outcomes

pubmed_search_articles( queryTerm="SGLT2 inhibitor cardiovascular outcomes randomized controlled trial", maxResults=20, sortBy="relevance" )
pubmed_search_articles( queryTerm="SGLT2 inhibitor cardiovascular outcomes randomized controlled trial", maxResults=20, sortBy="relevance" )

Then fetch full details for top results

Then fetch full details for top results

pubmed_fetch_contents(pmids=["PMID1", "PMID2", ...])
pubmed_fetch_contents(pmids=["PMID1", "PMID2", ...])

Find related articles for key papers

Find related articles for key papers

pubmed_article_connections( sourcePmid="key_paper_pmid", relationshipType="pubmed_similar_articles" )
undefined
pubmed_article_connections( sourcePmid="key_paper_pmid", relationshipType="pubmed_similar_articles" )
undefined

Parallel Execution Pattern

并行执行模式

For independent threads, execute PubMed searches in parallel (multiple tool calls in one turn), then write each to its thread file.
Example: Comparing SGLT2 Inhibitors
Thread 1: Empagliflozin → pubmed_search "empagliflozin cardiovascular RCT" → threads/empagliflozin.md
Thread 2: Dapagliflozin → pubmed_search "dapagliflozin cardiovascular RCT" → threads/dapagliflozin.md
Thread 3: Canagliflozin → pubmed_search "canagliflozin cardiovascular RCT" → threads/canagliflozin.md
Execute all three searches, then write findings to respective files.
对于相互独立的研究线程,并行执行 PubMed 搜索(在单轮中发起多个工具调用),然后将每个线程的结果写入对应的线程文件。
示例:比较 SGLT2 抑制剂
Thread 1: Empagliflozin → pubmed_search "empagliflozin cardiovascular RCT" → threads/empagliflozin.md
Thread 2: Dapagliflozin → pubmed_search "dapagliflozin cardiovascular RCT" → threads/dapagliflozin.md
Thread 3: Canagliflozin → pubmed_search "canagliflozin cardiovascular RCT" → threads/canagliflozin.md
执行全部三次搜索,然后将研究发现写入各自的文件。

Sequential Execution Pattern

顺序执行模式

For dependent threads, complete each fully before starting the next.
对于存在依赖关系的研究线程,必须先完整完成一个线程,再开始下一个。

Thread File Format

线程文件格式

Each
threads/thread_N.md
should contain:
markdown
undefined
每个
threads/thread_N.md
都应包含:
markdown
undefined

Thread: {Subtopic}

Thread: {Subtopic}

PubMed Searches Executed

PubMed Searches Executed

  1. Query: [exact query] → [N results] → Top PMIDs: [list]
  2. Query: [exact query] → [N results] → Top PMIDs: [list]
  1. Query: [exact query] → [N results] → Top PMIDs: [list]
  2. Query: [exact query] → [N results] → Top PMIDs: [list]

Key Findings

Key Findings

Finding 1: [Title]

Finding 1: [Title]

  • PMID: [number]
  • Citation: [Authors, Journal, Year]
  • Study type: RCT / Meta-analysis / Cohort / etc.
  • Population: [N patients, characteristics]
  • Key result: [HR/OR with 95% CI, p-value]
  • Quality: High / Medium / Low [+ brief justification]
  • PMID: [number]
  • Citation: [Authors, Journal, Year]
  • Study type: RCT / Meta-analysis / Cohort / etc.
  • Population: [N patients, characteristics]
  • Key result: [HR/OR with 95% CI, p-value]
  • Quality: High / Medium / Low [+ brief justification]

Finding 2: [Title]

Finding 2: [Title]

  • PMID: [number] ...
  • PMID: [number] ...

Contradictions Found

Contradictions Found

  • PMID X says [claim], PMID Y says [different claim]
  • Potential explanation: [patient population, endpoints, timing, etc.]
  • PMID X says [claim], PMID Y says [different claim]
  • Potential explanation: [patient population, endpoints, timing, etc.]

Gaps Identified

Gaps Identified

  • No RCT data on [specific question]
  • Limited evidence in [patient subgroup]
  • No RCT data on [specific question]
  • Limited evidence in [patient subgroup]

Thread Summary

Thread Summary

[2-3 sentence synthesis of this thread's findings with key PMIDs cited]
undefined
[2-3 sentence synthesis of this thread's findings with key PMIDs cited]
undefined

Context Offloading

上下文卸载

After every 5-7 tool calls:
  1. Write current findings to appropriate file
  2. Update
    progress.md
    with status
  3. Continue with fresh context
Trigger for offload:
  • Context feeling "full" (responses slowing, losing track)
  • Switching between threads
  • Before any synthesis step

每执行 5-7 次工具调用后:
  1. 将当前研究发现写入相应文件
  2. 使用最新状态更新
    progress.md
  3. 以新的上下文继续工作
触发卸载的条件:
  • 感觉上下文“已满”(响应变慢或开始失去线索)
  • 在研究线程之间切换时
  • 进入任何综合步骤之前

Step 4: Validate and Cross-Reference

第 4 步:验证与交叉引用

Read all thread files, then create
validation.md
:
markdown
undefined
读取所有线程文件,然后创建
validation.md
:
markdown
undefined

Validation Report

Validation Report

Facts Requiring Cross-Reference

Facts Requiring Cross-Reference

ClaimThread SourcePMIDVerification StatusConfidence
SGLT2i reduces HF hospitalizationThread 112345678Confirmed by PMIDs 23456789, 34567890High
Benefit extends to HFpEFThread 245678901Conflicting: PMID 56789012 shows nullInvestigate
ClaimThread SourcePMIDVerification StatusConfidence
SGLT2i reduces HF hospitalizationThread 112345678Confirmed by PMIDs 23456789, 34567890High
Benefit extends to HFpEFThread 245678901Conflicting: PMID 56789012 shows nullInvestigate

Contradictions Analysis

Contradictions Analysis

Contradiction 1: [Description]

Contradiction 1: [Description]

  • Position A: PMID [X], [study name], found [result]
  • Position B: PMID [Y], [study name], found [result]
  • Resolution: [Population difference / endpoint difference / timing / unresolved]
  • Position A: PMID [X], [study name], found [result]
  • Position B: PMID [Y], [study name], found [result]
  • Resolution: [Population difference / endpoint difference / timing / unresolved]

Source Quality Assessment

Source Quality Assessment

PMIDStudyTypeNQualityNotes
12345678EMPA-REGRCT7,020HighIndustry-funded but well-designed
23456789Meta-analysisMA45,000HighPublished in Lancet
PMIDStudyTypeNQualityNotes
12345678EMPA-REGRCT7,020HighIndustry-funded but well-designed
23456789Meta-analysisMA45,000HighPublished in Lancet

Validated Knowledge Base

Validated Knowledge Base

[List of facts we're confident in, with PMIDs]
  1. SGLT2 inhibitors reduce CV death in T2DM with established CVD (PMID: 12345678, 23456789)
  2. Benefit on HF hospitalization is consistent across the class (PMID: 34567890, 45678901)
  3. ...

---
[List of facts we're confident in, with PMIDs]
  1. SGLT2 inhibitors reduce CV death in T2DM with established CVD (PMID: 12345678, 23456789)
  2. Benefit on HF hospitalization is consistent across the class (PMID: 34567890, 45678901)
  3. ...

---

Step 5: Synthesize from Files

第 5 步:基于文件进行综合

Critical: Read from files, not memory.
bash
undefined
关键:从文件而非记忆中读取。
bash
undefined

Read all thread files

Read all thread files

cat ~/research_{topic}/threads/*.md
cat ~/research_{topic}/threads/*.md

Read validation

Read validation

cat ~/research_{topic}/validation.md

Write `synthesis.md`:

```markdown
cat ~/research_{topic}/validation.md

编写 `synthesis.md`:

```markdown

Synthesis: {Topic}

Synthesis: {Topic}

Cross-Thread Patterns

Cross-Thread Patterns

[What themes emerge across multiple threads?]
[What themes emerge across multiple threads?]

Key Insights

Key Insights

  1. [Insight that required combining multiple threads]
  2. [Insight that wasn't obvious in any single thread]
  3. ...
  1. [Insight that required combining multiple threads]
  2. [Insight that wasn't obvious in any single thread]
  3. ...

The Answer

The Answer

[Direct response to the primary research question, with PMID citations]
[Direct response to the primary research question, with PMID citations]

Evidence Strength Assessment

Evidence Strength Assessment

  • Strong evidence (multiple RCTs): [claims]
  • Moderate evidence (single RCT or consistent observational): [claims]
  • Limited evidence (observational only): [claims]
  • Expert opinion / guideline extrapolation: [claims]
  • Strong evidence (multiple RCTs): [claims]
  • Moderate evidence (single RCT or consistent observational): [claims]
  • Limited evidence (observational only): [claims]
  • Expert opinion / guideline extrapolation: [claims]

Remaining Gaps

Remaining Gaps

[What we still don't know and would need to investigate further]

---
[What we still don't know and would need to investigate further]

---

Step 6: Generate Final Report

第 6 步:生成最终报告

Write
report.md
using the synthesis:
markdown
undefined
基于综合结果编写
report.md
:
markdown
undefined

{Title}

{Title}

Executive Summary

Executive Summary

[3-5 sentences: question, key finding, main conclusion with strongest PMID]
[3-5 sentences: question, key finding, main conclusion with strongest PMID]

Research Question and Scope

Research Question and Scope

[From plan.md]
[From plan.md]

Methodology

Methodology

  • Database: PubMed via NCBI MCP
  • Search date: [date]
  • Total articles screened: [N]
  • Articles included: [N]
  • Study types: [breakdown]
  • Database: PubMed via NCBI MCP
  • Search date: [date]
  • Total articles screened: [N]
  • Articles included: [N]
  • Study types: [breakdown]

Findings

Findings

{Theme 1}

{Theme 1}

[Narrative synthesis with inline PMID citations]
[Narrative synthesis with inline PMID citations]

{Theme 2}

{Theme 2}

...
...

Analysis

Analysis

[Patterns, implications, connections]
[Patterns, implications, connections]

Conclusions

Conclusions

  1. [Primary conclusion with evidence level]
  2. [Secondary conclusions]
  1. [Primary conclusion with evidence level]
  2. [Secondary conclusions]

Clinical Implications

Clinical Implications

[If applicable: what this means for practice]
[If applicable: what this means for practice]

Limitations

Limitations

  • [Search limitations]
  • [Evidence gaps]
  • [Potential biases]
  • [Search limitations]
  • [Evidence gaps]
  • [Potential biases]

References

References

[Full reference list with PMIDs and DOIs]
  1. Author A, Author B, et al. Title. Journal. Year;Vol:Pages. PMID: XXXXXXXX. DOI: XX.XXXX/XXXXX
  2. ...

---
[Full reference list with PMIDs and DOIs]
  1. Author A, Author B, et al. Title. Journal. Year;Vol:Pages. PMID: XXXXXXXX. DOI: XX.XXXX/XXXXX
  2. ...

---

Parallel Research Patterns

并行研究模式

Pattern A: Drug/Entity Comparison

模式 A:药物或实体比较

Use when: Comparing 2-5 similar entities (drugs, devices, techniques)
User: "Compare CV outcomes of GLP-1 agonists"
→ Thread per drug (semaglutide, tirzepatide, liraglutide)
→ All threads parallel (same PubMed structure)
→ Comparison matrix synthesis
适用场景: 比较 2-5 个相似实体(药物、器械、技术)
User: "Compare CV outcomes of GLP-1 agonists"
→ Thread per drug (semaglutide, tirzepatide, liraglutide)
→ All threads parallel (same PubMed structure)
→ Comparison matrix synthesis

Pattern B: Pro/Con Analysis

模式 B:正反方分析

Use when: Topic has debate or controversy
User: "Analyze the evidence on aggressive LDL lowering"
→ Thread 1: Evidence FOR aggressive targets (PubMed: LDL <55 outcomes)
→ Thread 2: Evidence AGAINST/concerns (PubMed: LDL lowering adverse effects)
→ Thread 3: Current guidelines (fetch ACC/ESC guideline URLs)
→ Threads 1-2 parallel, Thread 3 after
适用场景: 主题存在辩论或争议
User: "Analyze the evidence on aggressive LDL lowering"
→ Thread 1: Evidence FOR aggressive targets (PubMed: LDL <55 outcomes)
→ Thread 2: Evidence AGAINST/concerns (PubMed: LDL lowering adverse effects)
→ Thread 3: Current guidelines (fetch ACC/ESC guideline URLs)
→ Threads 1-2 parallel, Thread 3 after

Pattern C: Evidence + Guidelines

模式 C:证据与指南

Use when: Need both primary evidence and clinical guidance
User: "What's the evidence on TAVR durability?"
→ Thread 1: Trial data (PubMed: TAVR long-term outcomes RCT)
→ Thread 2: Registry data (PubMed: TAVR registry durability)
→ Thread 3: Guidelines (fetch ACC/ESC valve guidelines)
→ All parallel
适用场景: 同时需要原始证据和临床指导
User: "What's the evidence on TAVR durability?"
→ Thread 1: Trial data (PubMed: TAVR long-term outcomes RCT)
→ Thread 2: Registry data (PubMed: TAVR registry durability)
→ Thread 3: Guidelines (fetch ACC/ESC valve guidelines)
→ All parallel

Pattern D: Historical Evolution

模式 D:历史演变

Use when: Understanding how evidence has evolved
User: "How has heart failure treatment evolved?"
→ Thread 1: Pre-neurohormonal era (PubMed: heart failure treatment 1980-1990)
→ Thread 2: ACE/ARB/BB era (PubMed: heart failure ACE inhibitor landmark)
→ Thread 3: Modern era ARNI/SGLT2 (PubMed: heart failure SGLT2 ARNI)
→ Sequential (each builds context for next)

适用场景: 了解证据如何随时间演变
User: "How has heart failure treatment evolved?"
→ Thread 1: Pre-neurohormonal era (PubMed: heart failure treatment 1980-1990)
→ Thread 2: ACE/ARB/BB era (PubMed: heart failure ACE inhibitor landmark)
→ Thread 3: Modern era ARNI/SGLT2 (PubMed: heart failure SGLT2 ARNI)
→ Sequential (each builds context for next)

Quality Checkpoints

质量检查点

After Step 2 (Planning)

第 2 步(规划)之后

  • Research question is specific and answerable
  • PubMed search strategies are defined for each thread
  • Threads are independent where marked parallel
  • Expected study types are specified
  • 研究问题明确且可以回答
  • 已为每个线程定义 PubMed 搜索策略
  • 标记为并行执行的线程相互独立
  • 已指定预期的研究类型

After Step 3 (Execution)

第 3 步(执行)之后

  • Each thread has 3+ credible PubMed sources
  • Key claims have specific data (HR, CI, p-value)
  • All citations have PMIDs
  • Gaps and contradictions are documented
  • Thread summaries are written
  • 每个线程包含至少 3 个可信的 PubMed 来源
  • 关键主张包含具体数据(HR、CI、p-value)
  • 所有引文均包含 PMID
  • 已记录证据缺口和矛盾
  • 已编写线程摘要

After Step 4 (Validation)

第 4 步(验证)之后

  • Key facts cross-referenced across threads
  • Contradictions analyzed with potential explanations
  • Source quality assessed for each major citation
  • Validated knowledge base compiled
  • 关键事实已在各线程之间进行交叉引用
  • 已分析矛盾及其潜在解释
  • 已评估每条主要引文的来源质量
  • 已汇编经过验证的知识库

After Step 5 (Synthesis)

第 5 步(综合)之后

  • Cross-thread patterns identified
  • Primary question directly answered
  • Evidence strength honestly assessed
  • Insights go beyond any single thread
  • 已识别跨线程模式
  • 已直接回答主要研究问题
  • 已客观评估证据强度
  • 洞见超越任何单个线程

Before Delivery

交付之前

  • Report structure matches user's requested format
  • All claims have PMID citations
  • Executive summary is truly executive (skimmable)
  • Reference list is complete with DOIs

  • 报告结构符合用户要求的格式
  • 所有主张均附有 PMID 引文
  • 执行摘要确实具备便于快速浏览的特点
  • 参考文献列表完整并包含 DOI

Common Research Pitfalls

常见研究误区

PitfallSymptomFix
Context overflowLosing track of earlier findingsWrite to files every 5-7 tool calls
Confirmation biasAll sources agree suspiciouslySearch for contradicting evidence explicitly
Recency biasOnly 2023-2024 sourcesInclude landmark trials regardless of date
Source homogeneityAll RCTs, no guidelinesAdd guideline thread for clinical context
Scope creepResearch expanding endlesslyReturn to plan.md, enforce boundaries
Premature synthesisConcluding before validationComplete Step 4 before Step 5
Memory-based synthesisCiting from recallRead files explicitly during Step 5
Non-PubMed citationsCiting Perplexity/webDelete and replace with PubMed source

误区表现修复方案
上下文溢出失去对早期发现的追踪每 5-7 次工具调用将结果写入文件
确认偏误所有来源异常一致明确搜索相互矛盾的证据
新近性偏误仅有 2023-2024 年的来源无论日期都纳入里程碑试验
来源同质化全部为 RCT,没有指南添加指南线程以补充临床背景
范围蔓延研究范围不断扩大返回
plan.md
,严格执行边界
过早综合在验证之前得出结论完成第 4 步后再执行第 5 步
基于记忆的综合凭记忆引用在第 5 步期间明确读取文件
非 PubMed 引文引用 Perplexity/Web删除并替换为 PubMed 来源

Example: Full Research Session

示例:完整研究会话

User: "Research the current evidence on colchicine for cardiovascular prevention"
Step 1: Initialize
bash
mkdir -p ~/research_colchicine_cv/sources
mkdir -p ~/research_colchicine_cv/threads
Step 2: Plan (write to plan.md)
  • Primary question: What's the evidence for colchicine in CV prevention?
  • Thread 1: Major RCTs (COLCOT, LoDoCo2, CLEAR SYNERGY)
    • PubMed: "colchicine cardiovascular randomized controlled trial"
  • Thread 2: Mechanisms and anti-inflammatory hypothesis
    • PubMed: "colchicine inflammation atherosclerosis mechanism"
  • Thread 3: Guidelines and clinical adoption
    • Fetch: ACC/ESC guideline URLs for stable CAD
  • Thread 4: Safety and practical considerations
    • PubMed: "colchicine adverse effects cardiovascular"
  • Threads 1, 2, 4 parallel; Thread 3 after 1 completes
Step 3: Execute
python
undefined
用户:“研究秋水仙碱用于心血管预防的当前证据”
第 1 步:初始化
bash
mkdir -p ~/research_colchicine_cv/sources
mkdir -p ~/research_colchicine_cv/threads
第 2 步:规划(写入
plan.md
)
  • 主要问题:秋水仙碱用于心血管预防的证据是什么?
  • 线程 1:主要 RCT(COLCOT、LoDoCo2、CLEAR SYNERGY)
    • PubMed:"colchicine cardiovascular randomized controlled trial"
  • 线程 2:机制与抗炎假说
    • PubMed:"colchicine inflammation atherosclerosis mechanism"
  • 线程 3:指南与临床采纳情况
    • 获取稳定型 CAD 的 ACC/ESC 指南 URL
  • 线程 4:安全性与实际考量
    • PubMed:"colchicine adverse effects cardiovascular"
  • 线程 1、2、4 可并行;线程 3 在线程 1 完成后执行
第 3 步:执行
python
undefined

Parallel searches

Parallel searches

pubmed_search_articles(queryTerm="colchicine cardiovascular randomized controlled trial", maxResults=15) pubmed_search_articles(queryTerm="colchicine inflammation atherosclerosis mechanism", maxResults=10) pubmed_search_articles(queryTerm="colchicine adverse effects cardiovascular", maxResults=10)
pubmed_search_articles(queryTerm="colchicine cardiovascular randomized controlled trial", maxResults=15) pubmed_search_articles(queryTerm="colchicine inflammation atherosclerosis mechanism", maxResults=10) pubmed_search_articles(queryTerm="colchicine adverse effects cardiovascular", maxResults=10)

Fetch top results

Fetch top results

pubmed_fetch_contents(pmids=["31733140", "32865377", "37634428"]) # COLCOT, LoDoCo2, CLEAR
pubmed_fetch_contents(pmids=["31733140", "32865377", "37634428"]) # COLCOT, LoDoCo2, CLEAR

Write to thread files

Write to thread files


**Step 4: Validate**
- Read all thread files
- Cross-reference mortality data across trials
- Note: CLEAR SYNERGY neutral vs positive COLCOT/LoDoCo2
- Analyze: Patient population differences (post-ACS vs chronic CAD)
- Write validation.md

**Step 5: Synthesize**
- Read from files
- Pattern: Inflammation hypothesis supported, but patient selection matters
- Insight: Post-ACS (COLCOT) benefit clear; chronic stable CAD (CLEAR) less certain
- Write synthesis.md

**Step 6: Report**
- Structured report with evidence summary
- Clear recommendation by patient type
- All PMIDs cited
- Complete reference list

---

**第 4 步:验证**
- 读取所有线程文件
- 交叉引用各试验中的死亡数据
- 注意:CLEAR SYNERGY 结果中性,而 COLCOT/LoDoCo2 结果阳性
- 分析患者人群差异(ACS 后与慢性 CAD)
- 写入 `validation.md`

**第 5 步:综合**
- 从文件读取
- 模式:炎症假说得到支持,但患者选择非常重要
- 洞见:ACS 后人群(COLCOT)的获益明确;慢性稳定型 CAD(CLEAR)的不确定性更高
- 写入 `synthesis.md`

**第 6 步:报告**
- 输出结构化报告并附证据摘要
- 按患者类型给出明确建议
- 引用所有 PMID
- 提供完整的参考文献列表

---

Integration with Other Skills

与其他 skill 的集成

This skill provides research foundation for:
  • cardiology-editorial
    → Use research output for trial analysis
  • cardiology-newsletter-writer
    → Research before writing
  • youtube-script-master
    → Research for script evidence base
  • x-post-creator-skill
    → Research before tweet generation
Workflow:
  1. User requests content on topic
  2. Run
    deep-researcher
    first (this skill)
  3. Pass validated findings to writing skill
  4. Writing skill cites PMIDs from research output

此 skill 为以下内容提供研究基础:
  • cardiology-editorial
    → 使用研究输出进行试验分析
  • cardiology-newsletter-writer
    → 写作前先开展研究
  • youtube-script-master
    → 为脚本提供循证依据
  • x-post-creator-skill
    → 生成推文前先开展研究
工作流:
  1. 用户请求某个主题的内容
  2. 首先运行
    deep-researcher
    (此 skill)
  3. 将经过验证的发现传递给写作 skill
  4. 写作 skill 引用研究输出中的 PMID

When NOT to Use This Skill

不应使用此 skill 的场景

  • Simple factual questions (use PubMed MCP directly)
  • Trend discovery (use Perplexity, but don't cite)
  • Non-medical topics (this skill is optimized for PubMed)
  • Quick content needs (use writing skill directly with inline research)
Use this skill when you need:
  • 5+ sources synthesized
  • Complex multi-faceted questions
  • Rigorous evidence assessment
  • Comprehensive literature coverage
  • 简单事实问题(直接使用 PubMed MCP)
  • 趋势发现(使用 Perplexity,但不要引用)
  • 非医学主题(此 skill 针对 PubMed 进行了优化)
  • 快速内容需求(直接使用写作 skill,并内嵌研究)
在以下情况下使用此 skill:
  • 需要综合 5 个以上来源
  • 存在复杂的多方面问题
  • 需要严格的证据评估
  • 需要全面的文献覆盖