Research Synthesis

When to Use This Skill

Use research-synthesis when:

• User mentions a claim that needs supporting data
• Need recent examples or trends ("what's happening with X lately?")
• Looking for citations or authoritative sources
• Extracting information from specific URLs
• Checking technical documentation or library APIs
• Filling gaps in knowledge during brainstorming or drafting

Skip when:

• Information is clearly from personal experience (no need to verify)
• User explicitly says "I don't need research, just write"
• Topic is purely opinion-based without factual claims

Critical: Never Hallucinate or Make Up Data

Only use REAL research from MCP tools. Never invent information.

NEVER make up:

• Statistics or percentages ("70% of companies...")
• Study names or researchers ("HBR study shows...")
• Company examples or case studies
• Technical specifications or benchmarks
• Quotes or citations

ALWAYS use MCP tools for research:

• Perplexity: For real data, studies, trends
• Firecrawl: For actual content from URLs
• Context7: For official documentation

If you can't find real data:

❌ BAD (hallucinating):
AI: Research shows 70% of OKR implementations fail...

✓ GOOD (admitting lack of data):
AI: I don't have data on OKR failure rates. Should I research this
    using Perplexity, or would you prefer to skip the statistic?

Before adding research to braindump:

• Verify it came from MCP tool results (not your training data)
• Include source attribution always
• If uncertain about accuracy, say so
• Don't "fill in" missing details with assumptions

MCP Tool Selection

Perplexity (Broad Research)

Use for:

• Recent trends, news, or developments ("what's the latest on X?")
• Statistical data or research findings
• Multiple perspectives on a topic
• Discovering examples or case studies
• General knowledge gaps

Examples:

• "Recent research on OKR implementation failures"
• "Companies that abandoned agile methodologies"
• "Studies on remote work productivity metrics"

How to invoke:

AI: Let me research recent examples...
    [uses Perplexity]
    Found several relevant data points - adding to braindump...

Firecrawl (Specific URL Extraction)

Use for:

• User mentions specific article or blog post
• Extracting content from known URLs
• Getting full text from referenced sources
• When user says "check out this article..."

Examples:

• User: "There's a good HBR article on this"
• User: "Add this to research: https://..."
• Extracting quotes from referenced articles

How to invoke:

AI: I'll extract the key points from that article...
    [uses Firecrawl]
    Here's what I found - should I add this to braindump?

Context7 (Technical Documentation)

Use for:

• Library or framework documentation
• API references
• Technical specifications
• When writing about specific technologies

Examples:

• "How does React's useEffect actually work?"
• "What are the official recommendations for X?"
• "Check the latest API docs for Y"

How to invoke:

AI: Let me check the official docs...
    [uses Context7]
    According to the documentation...

Decision Tree

Need research?
├─ Specific URL mentioned? → Firecrawl
├─ Technical docs/APIs? → Context7
└─ Everything else → Perplexity

Synthesizing Findings

Don't Just Dump Data

Bad:

Research shows:
- Stat 1
- Stat 2
- Stat 3
[Added to braindump]

Good:

Found interesting pattern: 3 recent studies show 60-70% of OKR
implementations fail to improve performance. Notable:
- HBR research: 70% failure rate, primarily due to metric gaming
- McKinsey: Companies with >100 OKRs see diminishing returns
- Google's evolution: Shifted from strict OKRs to "goals and signals"

Key insight: Failure correlates with treating OKRs as compliance
exercise rather than alignment tool.

[Added to braindump under Research]

Synthesize Patterns, Not Lists

Look for:

• Patterns: What do multiple sources agree on?
• Contradictions: Where do sources disagree?
• Gaps: What's missing from the research?
• Surprises: What's unexpected or counterintuitive?

Maintain Source Attribution

Format for citations in braindump:

## Research

### OKR Implementation Failures
Studies show 60-70% failure rate (HBR, McKinsey).
Primary causes: metric gaming, checkbox compliance.

**Sources:**
- HBR: "Why OKRs Don't Work" - 70% fail to improve performance
- McKinsey: Survey of 500 companies, correlation with OKR count
- Google's blog: Evolution of their goals system over recent years

**Key Quote:**
> "When OKRs become a performance evaluation tool, they stop being
> a planning tool." - John Doerr, Measure What Matters

Integration with Conversation

Research should flow naturally, not interrupt:

Proactive Research

You: I think OKRs fail because of Goodhart's Law

AI: That's a strong claim - let me check recent data...
    [uses Perplexity]

    Good intuition! Found 3 studies that confirm this:
    [synthesized findings]

    Adding this to braindump under Research. Want to explore
    why Goodhart's Law applies specifically to OKRs?

User-Requested Research

You: Find examples of companies that abandoned OKRs

AI: [uses Perplexity]
    Found several notable cases:
    - Company X: Switched to narrative planning (Amazon style)
    - Company Y: Reverted to simpler quarterly goals
    - Google: Evolved to "Goals and Signals" framework

    Should I add all of these to braindump, or focus on
    one specific approach?

During Drafting

You: I need a citation for this claim

AI: [uses Perplexity]
    Found supporting research:
    [citation details]

    I'll add this to the draft with proper attribution.

Adding to Braindump

Always ask before updating braindump.md (unless context is clear):

AI: I found X, Y, and Z. Should I add this to braindump
    under Research, or would you like to review first?

Update structured sections:

• Research: Studies, data, citations
• Examples: Concrete cases and stories
• Quotes: Notable quotations with attribution
• Sources: Full references for later citation

Quality Standards

Before adding research to braindump:

• Synthesized into narrative, not just listed
• Source attribution included
• Relevance to core argument is clear
• Key insights or patterns highlighted
• Contradictions or gaps noted if relevant

Common Pitfalls to Avoid

1. Information Overload: Don't dump 20 sources - synthesize 3-5 key findings
2. Missing Attribution: Always cite sources for later reference
3. Irrelevant Research: Just because you found it doesn't mean it's useful
4. Breaking Flow: Don't interrupt creative flow for minor fact-checks
5. Uncritical Acceptance: Note when sources disagree or have limitations

Example Flow

For detailed conversation examples showing research synthesis techniques and MCP tool usage, see reference/examples.md.

Integration with Other Skills

• During brainstorming: Research validates or challenges initial ideas
• During drafting (blog-writing): Research provides citations and examples
• Throughout: Update braindump.md with structured findings