# Research Topic

> Conduct deep research on any topic using a structured multi-angle investigation framework. Use this skill when asked to research something, investigate a topic, do a deep dive, compile a report, or gather comprehensive information on any subject. Triggers include "research", "investigate", "deep dive", "report on", "analyze", "what do we know about", "comprehensive overview", or requests for thorough analysis of any topic.

- Skill: `freeman983/research-topic` (Agent Skill)
- Install (CLI): `npx skillmds@latest add freeman983/research-topic`
- Raw SKILL.md: https://api.skillmd.com/api/skills/freeman983/research-topic/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Docs & Writing
- Author: freeman983 (https://skillmd.com/u/freeman983)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/freeman983/research-topic

---


# Deep Topic Research

Conduct structured research that produces actionable intelligence, not just information dumps. The goal is synthesis — connecting dots that aren't obvious.

## Core Method: The Diamond Framework

```
        Diverge (Gather)
       /                \
      /                  \
     /    Wide Search     \
    /                      \
   /________________________\
   \                        /
    \   Filter & Verify    /
     \                    /
      \   Synthesize     /
       \                /
        Converge (Insight)
```

1. **Diverge** — Cast a wide net across multiple angles
2. **Filter** — Separate signal from noise, verify claims
3. **Synthesize** — Connect patterns, identify insights
4. **Converge** — Deliver clear conclusions with confidence levels

## Research Angles

For any topic, investigate from these 6 angles:

| Angle | Questions | Sources |
|-------|-----------|---------|
| **What** | Definition, scope, current state | Official docs, Wikipedia, industry reports |
| **Why** | Root causes, motivations, drivers | Academic papers, expert interviews, history |
| **How** | Mechanisms, processes, implementation | Technical docs, case studies, tutorials |
| **Who** | Key players, stakeholders, experts | LinkedIn, company pages, conference talks |
| **When** | Timeline, trends, milestones | News archives, historical data, roadmaps |
| **So What** | Implications, predictions, actions | Analysis pieces, expert opinions, data trends |

## Research Workflow

### Phase 1: Scope Definition
Before searching, define:
- **Core question**: What exactly are we trying to answer?
- **Depth level**: Overview / Working knowledge / Expert analysis
- **Time frame**: Current state / Historical / Forward-looking
- **Audience**: Who will read this and what decisions will they make?

### Phase 2: Wide Search
Run parallel searches across multiple dimensions:

```
Primary searches (run in parallel):
- "[Topic] overview 2025"
- "[Topic] latest developments"
- "[Topic] research papers"
- "[Topic] expert analysis"
- "[Topic] criticism challenges"
- "[Topic] vs alternatives comparison"
- "[Topic] future predictions trends"
```

### Phase 3: Source Evaluation

Rate each source on credibility:

| Tier | Source Type | Trust Level |
|------|-----------|-------------|
| **Tier 1** | Peer-reviewed papers, official data, primary sources | High — cite directly |
| **Tier 2** | Reputable journalism, industry reports, expert blogs | Medium-High — cite with attribution |
| **Tier 3** | News aggregators, Wikipedia, general articles | Medium — verify claims elsewhere |
| **Tier 4** | Social media, forums, anonymous sources | Low — use for leads only, never cite alone |

### Phase 4: Cross-Reference
- Every major claim needs 2+ independent sources
- Flag single-source claims as [UNVERIFIED]
- Note contradictions between sources explicitly
- Distinguish between facts, expert opinions, and speculation

### Phase 5: Synthesis
- Identify patterns across sources
- Note what multiple experts agree on (consensus)
- Note where experts disagree (open questions)
- Connect to adjacent topics the user might not have considered
- Form original insights from the combination of sources

## Language Rules

- Match the language of the user's request by default
- If the user explicitly specifies a language (e.g., "in English", "用中文"), use that language for all output
- Keep proper nouns, brand names, and technical terms in their original language

## Output Structure

### Executive Summary Format (Default)

```markdown
# Research Report: [Topic]

## Executive Summary
[3-5 sentences: What this is, why it matters, key finding]

## Key Findings

### 1. [Finding Title]
[2-3 paragraphs with evidence]
- **Evidence**: [Data/source]
- **Confidence**: High / Medium / Low
- **Implication**: [So what?]

### 2. [Finding Title]
...

### 3. [Finding Title]
...

## Landscape Analysis
[Who are the key players? What are the competing approaches?]

## Open Questions
- [What we still don't know]
- [Where experts disagree]
- [What needs more investigation]

## Timeline
- [Key historical milestones]
- [Recent developments]
- [Expected future events]

## Recommendations
[Based on findings, what should the reader do?]

## Sources
[Numbered list with URLs, organized by tier]

---
📊 Research depth: [Overview / Working Knowledge / Expert]
🔍 Sources consulted: [N]
⚠️ Confidence level: [High / Medium / Low]
```

### Comparison Format (for "X vs Y" questions)

```markdown
# [X] vs [Y]: Research Report

## Quick Verdict
[One paragraph: which is better for what use case]

## Comparison Matrix

| Dimension | [X] | [Y] | Winner |
|-----------|-----|-----|--------|
| [Dim 1]   |     |     |        |
| [Dim 2]   |     |     |        |
...

## Detailed Analysis
### [Dimension 1]
[Deep comparison with evidence]

## When to Choose [X]
[Specific scenarios]

## When to Choose [Y]
[Specific scenarios]
```

## Confidence Calibration

Be explicit about certainty levels:

| Level | Meaning | Language |
|-------|---------|---------|
| **High** | Multiple Tier 1-2 sources agree | "Evidence strongly suggests..." |
| **Medium** | Some evidence, some gaps | "Available data indicates..." |
| **Low** | Limited or conflicting evidence | "Early signals suggest, but..." |
| **Speculative** | Informed guess, no hard data | "One possible interpretation..." |

## Research Quality Rules

1. **No unsourced claims** — every factual statement links to a source
2. **Distinguish fact from opinion** — "X happened" vs "Expert Y believes..."
3. **Include counter-arguments** — steelman the opposing view
4. **Recency matters** — flag if best available data is outdated
5. **Quantify when possible** — "significant growth" → "47% YoY growth"
6. **Acknowledge gaps** — what you couldn't find is as important as what you found

## Anti-Patterns

- **Confirmation bias** — only searching for evidence that supports a hypothesis
- **Source laundering** — multiple articles citing the same original source ≠ independent verification
- **Recency bias** — assuming the latest article is the most accurate
- **Authority bias** — big company said it ≠ it's true
- **Completeness theater** — listing 50 sources when 10 high-quality ones would be better
- **Analysis paralysis** — researching forever instead of delivering actionable findings

## Quick Checklist

1. Is the core question clearly defined?
2. Did I search from multiple angles (not just one framing)?
3. Are major claims supported by 2+ independent sources?
4. Did I include counter-arguments and open questions?
5. Is every finding tagged with a confidence level?
6. Does the report end with actionable recommendations?
7. Would someone make a better decision after reading this?