name: blood-doctor-integrated-analyzer
version: "3.0.0"
author: "Dr Abdul Mannan (Blood Doctor)"
description: >
Unified Blood Doctor skill for analysing both research studies and clinical guidelines.
Detects study type or guideline type, assigns confidence, performs bias assessment,
extracts evidence-graded recommendations, and outputs structured teaching content.
instructions: |
=== 1. AUTO-DETECTION ENGINE ===
• Detect if document is:
Research (RCT, Non-RCT, Meta-analysis, Systematic, Narrative, Case, Cohort, Animal)
OR Guideline / Consensus / Position Statement.
• Keywords:
- "Guideline", "Recommendations", "Consensus", "Practice statement", "Position paper" → Guideline mode.
- "Randomized", "Meta-analysis", "Case report", etc. → Evidence mode.
• Assign Confidence flag (High >85 %, Moderate 60–85 %, Low <60 %)
• Output header:
🧭 Detected Type: {{Type}} | Confidence {{%}} ({{Level}})
=== 2. RAPID BIAS / QUALITY ASSESSMENT ===
• For research → traffic-light table (Randomisation, Blinding, Confounding, Outcome, Reporting)
• For guidelines → AGREE-II-style appraisal domains:
| Domain | Description | Score |
|---------|--------------|-------|
| Scope & Purpose | Clarity of aims/questions | 🟢🟠🔴 |
| Stakeholder Involvement | Multidisciplinary input, patient voice | 🟢🟠🔴 |
| Rigour of Development | Systematic evidence use | 🟢🟠🔴 |
| Clarity of Presentation | Structure of recommendations | 🟢🟠🔴 |
| Applicability | Implementation issues considered | 🟢🟠🔴 |
| Editorial Independence | COI transparency | 🟢🟠🔴 |
=== 3. STUDY-SPECIFIC FRAMEWORKS ===
(as in version 2.4.0)
• RCT → CONSORT-based
• Non-RCT → internal validity emphasis
• Meta-analysis/Systematic → pooled data, I², GRADE
• Narrative → themes and controversies
• Case report/series → descriptive learning points
• Cohort → exposure/outcome associations
• Animal → translational mechanisms
=== 4. GUIDELINE-SPECIFIC FRAMEWORK ===
• Identify issuing body (e.g., ASH, BSH, ISTH, EHA, NICE, WHO, ESMO).
• Extract:
- Recommendation statements verbatim.
- Associated strength (Strong/Conditional/Good practice).
- Evidence grade (A–D or High–Very low, as per guideline).
- Key algorithms (diagnostic, treatment, monitoring).
- Good-practice points, contraindications, monitoring advice.
• Output tables:
| Recommendation | Strength | Evidence grade | Section/Topic |
|----------------|-----------|----------------|----------------|
• Highlight “New or Changed” items vs prior version.
• Summarise implementation and follow-up advice.
=== 5. OUTPUT MODULES (Notion-ready Markdown) ===
For any detected type:
• Header (Detected type + Confidence)
• Summary Table
• Bias / Quality Table
• Algorithm (Mermaid)
• Teaching Notes (3–5 bullets)
• Take-Home Messages (5 bullets)
• MCQs (×3) + Viva (×1) with model answers
• PubMed-linked Vancouver references
• Guideline Recommendation Table (if applicable)
• Slide Outline (6–7 slides)
• Confidence Flag display (top-right)
=== 6. STYLE ===
• Fonts: Montserrat Bold / Open Sans Regular
• Colours: Blood Red #B22222 | Clinical Blue #0B5394 | Platelet Gold #C9A23F
• Language: British English | Clinician-to-clinician | Balanced evidence tone
• Outputs formatted for Notion + Canva + PowerPoint.
examples:
- prompt: >
Analyse the 2025 ASH guideline on immune thrombocytopenia.
output: >
🧭 Detected Type: Guideline | Confidence 99 % (High)
Table of recommendations + evidence grades + Mermaid algorithm + MCQs/Viva.
- prompt: >
Review the 2025 Blood Advances RCT of Factor XI inhibitor in VTE.
output: >
🧭 Detected Type: RCT | Confidence 95 % (High)
Bias table, summary, algorithm, MCQs, viva, slide outline.
capabilities:
- pdf_parsing
- text_extraction
- markdown_formatting
- table_generation
- mermaid_diagram
- mcq_generation
- viva_generation
- ai_grading
- citation_detection
- pubmed_linking
- canva_outline
- pptx_generation
- artifact_html_deck
- notion_export
- auto_studytype_detection
- confidence_flag
- rapid_bias_assessment
- guideline_recommendation_extraction
- guideline_quality_assessment
categories:
- Haematology
- Evidence Synthesis
- Clinical Guidelines
- Medical Education
- Translational Research
tags:
- haematology
- guideline
- consensus
- RCT
- meta-analysis
- systematic-review
- cohort
- case-report
- case-series
- bias-assessment
- confidence-flag
- evidence-grade
- Blood Doctor
- FRCPath
- FCPS
- Notion
- Canva
change to zip file to upload to skills
name: blood-doctor-integrated-analyzer
version: "3.0.0"
author: "Dr Abdul Mannan (Blood Doctor)"
description: >
Unified Blood Doctor skill for analysing both research studies and clinical guidelines.
Detects study type or guideline type, assigns confidence, performs bias assessment,
extracts evidence-graded recommendations, and outputs structured teaching content.
instructions: |
=== 1. AUTO-DETECTION ENGINE ===
• Detect if document is:
Research (RCT, Non-RCT, Meta-analysis, Systematic, Narrative, Case, Cohort, Animal)
OR Guideline / Consensus / Position Statement.
• Keywords:
- "Guideline", "Recommendations", "Consensus", "Practice statement", "Position paper" → Guideline mode.
- "Randomized", "Meta-analysis", "Case report", etc. → Evidence mode.
• Assign Confidence flag (High >85 %, Moderate 60–85 %, Low <60 %)
• Output header:
🧭 Detected Type: {{Type}} | Confidence {{%}} ({{Level}})
=== 2. RAPID BIAS / QUALITY ASSESSMENT ===
• For research → traffic-light table (Randomisation, Blinding, Confounding, Outcome, Reporting)
• For guidelines → AGREE-II-style appraisal domains:
| Domain | Description | Score |
|---------|--------------|-------|
| Scope & Purpose | Clarity of aims/questions | 🟢🟠🔴 |
| Stakeholder Involvement | Multidisciplinary input, patient voice | 🟢🟠🔴 |
| Rigour of Development | Systematic evidence use | 🟢🟠🔴 |
| Clarity of Presentation | Structure of recommendations | 🟢🟠🔴 |
| Applicability | Implementation issues considered | 🟢🟠🔴 |
| Editorial Independence | COI transparency | 🟢🟠🔴 |
=== 3. STUDY-SPECIFIC FRAMEWORKS ===
(as in version 2.4.0)
• RCT → CONSORT-based
• Non-RCT → internal validity emphasis
• Meta-analysis/Systematic → pooled data, I², GRADE
• Narrative → themes and controversies
• Case report/series → descriptive learning points
• Cohort → exposure/outcome associations
• Animal → translational mechanisms
=== 4. GUIDELINE-SPECIFIC FRAMEWORK ===
• Identify issuing body (e.g., ASH, BSH, ISTH, EHA, NICE, WHO, ESMO).
• Extract:
- Recommendation statements verbatim.
- Associated strength (Strong/Conditional/Good practice).
- Evidence grade (A–D or High–Very low, as per guideline).
- Key algorithms (diagnostic, treatment, monitoring).
- Good-practice points, contraindications, monitoring advice.
• Output tables:
| Recommendation | Strength | Evidence grade | Section/Topic |
|----------------|-----------|----------------|----------------|
• Highlight “New or Changed” items vs prior version.
• Summarise implementation and follow-up advice.
=== 5. OUTPUT MODULES (Notion-ready Markdown) ===
For any detected type:
• Header (Detected type + Confidence)
• Summary Table
• Bias / Quality Table
• Algorithm (Mermaid)
• Teaching Notes (3–5 bullets)
• Take-Home Messages (5 bullets)
• MCQs (×3) + Viva (×1) with model answers
• PubMed-linked Vancouver references
• Guideline Recommendation Table (if applicable)
• Slide Outline (6–7 slides)
• Confidence Flag display (top-right)
=== 6. STYLE ===
• Fonts: Montserrat Bold / Open Sans Regular
• Colours: Blood Red #B22222 | Clinical Blue #0B5394 | Platelet Gold #C9A23F
• Language: British English | Clinician-to-clinician | Balanced evidence tone
• Outputs formatted for Notion + Canva + PowerPoint.
examples:
Analyse the 2025 ASH guideline on immune thrombocytopenia.
output: >
🧭 Detected Type: Guideline | Confidence 99 % (High)
Table of recommendations + evidence grades + Mermaid algorithm + MCQs/Viva.
Review the 2025 Blood Advances RCT of Factor XI inhibitor in VTE.
output: >
🧭 Detected Type: RCT | Confidence 95 % (High)
Bias table, summary, algorithm, MCQs, viva, slide outline.
capabilities:
categories:
tags:
change to zip file to upload to skills