From example-skills
Guides systematic research synthesis: scope questions, gather/evaluate sources, extract insights, synthesize themes, and produce reports or decision frameworks.
How this skill is triggered — by the user, by Claude, or both
Slash command
/example-skills:research-synthesis-workflowThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
This skill provides a systematic methodology for conducting research, synthesizing findings from multiple sources, and producing actionable knowledge artifacts.
This skill provides a systematic methodology for conducting research, synthesizing findings from multiple sources, and producing actionable knowledge artifacts.
┌──────────────────────────────────────────────────────────────┐
│ Research Synthesis Workflow │
├──────────────────────────────────────────────────────────────┤
│ │
│ 1. SCOPE 2. GATHER 3. EXTRACT │
│ ┌─────────┐ ┌─────────┐ ┌─────────┐ │
│ │ Define │─────▶│ Find │─────▶│ Capture │ │
│ │ Question│ │ Sources │ │ Insights│ │
│ └─────────┘ └─────────┘ └─────────┘ │
│ │ │ │
│ │ 5. PRODUCE 4. SYNTHESIZE │
│ │ ┌─────────┐ ┌─────────┐ │
│ └─────────▶│ Create │◀─────│ Connect │ │
│ │ Artifact│ │ Themes │ │
│ └─────────┘ └─────────┘ │
│ │
└──────────────────────────────────────────────────────────────┘
Transform vague topics into answerable questions:
| Type | Pattern | Example |
|---|---|---|
| Exploratory | What is X? How does X work? | What is vector search? |
| Comparative | How does X compare to Y? | PostgreSQL vs. Neo4j for graphs? |
| Evaluative | Is X effective for Y? | Is RAG effective for technical docs? |
| Causal | What causes X? What are effects of X? | What causes LLM hallucinations? |
| Prescriptive | How should we implement X? | How to design a RAG pipeline? |
Define explicitly:
## Research Scope: Vector Database Selection
### Research Question
Which vector database best fits our production RAG system
requiring <50ms latency at 10M+ vectors?
### In Scope
- Pinecone, Weaviate, Milvus, Qdrant, pgvector
- Latency benchmarks at scale
- Cost analysis (cloud vs self-hosted)
- Operational complexity
### Out of Scope
- General-purpose databases with vector extensions
- Sub-million vector use cases
- Academic/research-only systems
### Success Criteria
Recommendation with supporting evidence for 2-3 top candidates
Evaluate each source on:
| Criterion | High Quality | Low Quality |
|---|---|---|
| Authority | Expert author, peer-reviewed | Anonymous, no credentials |
| Currency | Recent, updated | Outdated, no dates |
| Accuracy | Citations, verifiable | Unsupported claims |
| Purpose | Inform, educate | Sell, persuade |
| Coverage | Comprehensive | Superficial |
Primary Sources (original)
├── Research papers
├── Official documentation
├── Benchmark data
└── Expert interviews
Secondary Sources (analysis)
├── Review articles
├── Technical blogs
├── Industry reports
└── Book chapters
Tertiary Sources (summaries)
├── Wikipedia
├── Textbooks
└── Encyclopedias
Keyword expansion:
Citation chaining:
Author tracking:
For each source, capture:
## Source: [Title]
- **URL/DOI**:
- **Author(s)**:
- **Date**:
- **Type**: [paper/blog/docs/report]
- **Quality Score**: [1-5]
- **Relevance**: [high/medium/low]
- **Key Topics**:
- **Notes**:
Use consistent templates for extraction:
## Claim: [Specific assertion]
- **Source**: [reference]
- **Evidence**: [supporting data/reasoning]
- **Strength**: [strong/moderate/weak]
- **My Assessment**: [agree/disagree/uncertain]
- **Related Claims**: [links to other notes]
| Type | Description | Weight |
|---|---|---|
| Empirical | Measured data, experiments | High |
| Analytical | Logical derivation | Medium-High |
| Anecdotal | Case studies, examples | Medium |
| Expert Opinion | Authority statements | Medium |
| Theoretical | Model predictions | Medium-Low |
When sources disagree:
## Conflict: [Topic]
### Position A: [Claim]
- Sources: [list]
- Evidence: [summary]
### Position B: [Claim]
- Sources: [list]
- Evidence: [summary]
### Analysis
- Methodological differences:
- Context differences:
- Possible resolution:
- My conclusion:
Codes Themes Findings
├─ fast queries ─┐
├─ low latency ─┼── Performance ─┬── Theme 1: Performance
├─ high throughput ─┘ │ varies significantly
├─ managed service ─┐ │ by workload type
├─ self-hosted ─┼── Deployment ─┼── Theme 2: Cloud vs
├─ kubernetes ─┘ │ self-hosted tradeoff
├─ pricing tiers ─┐ │
├─ compute costs ─┼── Economics ─┴── Theme 3: Total cost
├─ hidden costs ─┘ drives final choice
Create decision frameworks from synthesis:
## Vector Database Selection Framework
### Decision Tree
1. Scale requirement?
- <1M vectors → pgvector (simplicity)
- 1M-100M vectors → Continue to 2
- >100M vectors → Milvus/Weaviate (distributed)
2. Operational capacity?
- Limited DevOps → Pinecone (managed)
- Strong DevOps → Continue to 3
3. Cost sensitivity?
- Budget constrained → Qdrant (open source)
- Budget flexible → Evaluate all options
### Comparison Matrix
| Criterion | Weight | Pinecone | Milvus | Qdrant |
|----------------|--------|----------|--------|--------|
| Latency | 30% | 4 | 5 | 4 |
| Scalability | 25% | 5 | 5 | 4 |
| Operations | 20% | 5 | 3 | 4 |
| Cost | 15% | 2 | 4 | 5 |
| Features | 10% | 4 | 5 | 4 |
| **Weighted** | | **4.0** | **4.4**| **4.2**|
| Format | Purpose | Audience |
|---|---|---|
| Executive Summary | Quick decision support | Leadership |
| Technical Report | Detailed analysis | Engineers |
| Literature Review | Academic synthesis | Researchers |
| Decision Framework | Structured evaluation | Decision makers |
| Reference Guide | Quick lookup | Practitioners |
Executive Summary (1-2 pages):
Technical Report (5-20 pages):
Before finalizing:
Research is rarely linear:
references/evaluation-rubrics.md - Source quality scoring guidesreferences/synthesis-methods.md - Detailed synthesis techniquesreferences/artifact-templates.md - Document templates and examplesnpx claudepluginhub a-organvm/a-i--skills --plugin document-skillsProvides a checklist for code reviews covering functionality, security, performance, maintainability, tests, and quality. Use for pull requests, audits, team standards, and developer training.