Source: https://github.com/aipoch/medical-research-skills

Output Format

Output must strictly follow this structured report format to ensure users receive directly readable decision-ready content rather than a simple text description.

1. Hotspot Overview Table

Display all identified hotspot topics and their popularity metrics in tabular form.

• Popularity Index: Composite score based on publication volume, citation frequency, and top-tier journal proportion (max ★★★★★).
• Trend Direction: Recent 3-year publication trend (Rising / Stable / Declining).
• Representative Papers: Number of core papers matching the hotspot topic.
• Active Years: Year range with sustained output for the hotspot.

2. Hotspot Detail Analysis

Each hotspot is expanded independently, including popularity rating, trend description, sub-direction composition, and key papers.

Hotspot 1: [Topic Name]

• Popularity Index: ★★★★★ (Very Hot)
• Trend: Rising publication volume over the past 3 years
• Core Research Directions:
  • Sub-direction A (40%)
  • Sub-direction B (35%)
• Key Papers:

• Evidence Level: High (top-tier/highly cited), Medium (mainstream/moderate citations), Low (lower-tier/few citations).

Hotspot 2: [Topic Name]

...

3. Research Gap Identification

Analyze under-explored areas in current literature.

• Potential Value: High / Medium / Low, based on unmet clinical or basic research needs.
• Feasibility: High / Medium / Low, based on technical maturity, research barriers, and execution difficulty.

4. Recommended Entry Directions

Based on the preceding analysis, provide concrete actionable research entry suggestions.

---

Research Hotspot Analysis

When to Use

• The user provides a disease name, target, technical roadmap, or research topic, and wants to quickly see current research hotspots.
• The user needs to cluster recent literature by keywords and topics to find directions worth deeper exploration.
• The user wants a Markdown hotspot analysis report with representative literature for topic selection or review writing.

When Not to Use

• Do not use this skill when the user only needs single-paper retrieval or a simple reference list.
• If there is no clear disease, topic, or search scope, do not start clustering immediately — first ask the user to clarify topic boundaries.
• If the environment cannot access the scripts or retrieval data this skill depends on, do not fabricate hotspot results.

Required Inputs

Workflow

1. Use search_pubmed from scripts/analysis_ops.py to search relevant literature and obtain PMIDs and basic metadata.
2. Run word_frequency on the returned medline_texts to count high-frequency keywords or MeSH terms.
3. Combine with hotspot prompts in references/prompt_templates.md to cluster high-frequency keywords into 3-6 hotspot topics.
4. Use match_keywords to map representative literature to each topic, avoiding mismatches between topics and evidence.
5. For each topic, call sort_by_jif_and_select to choose representative literature, then use fetchPMCArticleDetails or fetchPubmedArticleDetails to supplement details.
6. Output a Markdown report with at least: research overview, hotspot topics, representative keywords per topic, representative literature, and follow-up suggestions.

Output Contract

• Primary output: A Markdown hotspot analysis report.
• Required fields: topic overview, hotspot topics, supporting papers, next-step suggestions.
• Recommend at least 2-3 representative papers per hotspot, with explanation of why the topic qualifies as a hotspot.
• If retrieval coverage is insufficient, must explicitly mark as PARTIAL.

Failure Handling

• Too few literature search results: First broaden time range or relax keywords, then explain coverage gaps.
• Unstable keyword clustering: Show high-frequency keywords and indicate clustering is candidate-only — do not force conclusions.
• Representative literature lacks usable details: Keep PMID and title, mark as pending.

User Checkpoints

• Before starting a broad search, confirm topic boundaries and time range.
• Before outputting the final hotspot report, if cluster topics are clearly ambiguous, send candidate topics to user for confirmation.

Tools

• fetchPMCArticleDetails: Get article details.
• fetchPubmedArticleDetails: Get PubMed details.

Scripts

• scripts/analysis_ops.py: Contains helper functions for PubMed search, frequency analysis, keyword matching, and result formatting.

References

• references/prompt_templates.md: Contains the system prompts for LLM analysis.

Input Validation

This skill accepts requests that match the documented purpose of research-hotspot-analysis and include enough context to complete the workflow safely.

Quick Validation

• Check that scripts/analysis_ops.py exists and can perform at least the three core steps: search, word frequency, and matching.
• Check that the final report contains hotspot topics with corresponding representative literature, not just a keyword list.
• Check that each hotspot topic has clear evidence sources to support it.