Research Agent

When to use

Use this skill for:

• market scans
• business idea research
• opportunity discovery
• competitor review
• customer-problem signal gathering
• thesis validation with public sources

Do not use this skill for:

• private-environment discovery
• legal or compliance advice
• deep technical implementation work
• pure coding tasks

Core outcome

Produce decision-ready research that:

• separates fact from inference
• logs sources clearly
• ranks opportunities instead of listing them loosely
• states uncertainty and open questions explicitly

Minimal graph

flowchart TD
    U([User Task]) --> B[Research Brief]
    B --> M[Market Scan]
    B --> C[Customer-Problem Scan]
    B --> K[Competitor Scan]
    M --> S[Synthesis]
    C --> S
    K --> S
    S --> R[Critique]
    R --> D([Deliver])

Stages

1. Research Brief

• define the question to answer
• define scope and exclusions
• define target customer and geography if relevant
• define what a useful output looks like

2. Market Scan

• collect public evidence of demand, tool adoption, category formation, or workflow pain
• prefer public sources with visible recency

3. Customer-Problem Scan

• look for direct user pain, repeated complaints, budget sensitivity, and operational friction
• prefer first-person reports over polished marketing claims

4. Competitor Scan

• identify substitutes, direct competitors, and adjacent approaches
• capture how the market is framed, priced, and differentiated

5. Synthesis

• score or rank the best opportunities
• explain why the recommendation wins over alternatives
• keep the recommendation narrow and actionable

6. Critique

• look for weak evidence
• note blocked or missing sources
• state what would disconfirm the conclusion

Evidence rules

• Separate facts, inference, and opinion.
• Date claims when recency matters.
• Log sources for every major conclusion.
• Prefer public, accessible sources.
• If key sources are blocked, say so clearly.
• Do not imply precise market size without reliable evidence.

Output contract

Every deliverable should include:

• objective
• recommendation
• why it wins over alternatives
• target customer
• risks
• open questions
• source log

Templates

• templates/research-brief.md
• templates/opportunity-report.md
• templates/source-log.md
• examples/example-opportunity-report.md

Rule of thumb

• Evidence first.
• Recommendation second.
• Hype never.

Example Opportunity Report

Executive summary

The best near-term opportunity is a productized service that converts public category and competitor research into a fixed-scope audit. It wins because it has low setup cost, strong pain visibility, and simple delivery.

Market context

Public category pages and founder discussions show rising interest in AI-assisted discovery and lightweight automation.

Customer problem signals

• founders say existing agencies are expensive and vague
• buyers want measurable improvements
• teams want low-input services, not long workshops

Alternatives and competitors

• generic SEO agencies
• outbound lead-gen agencies
• broad automation consultancies

Opportunity hypotheses

1. A narrow audit offer sells faster than a broad retainer. 2. Buyers respond better to evidence-backed remediation than to generic strategy decks.

Scoring matrix

Risks and disconfirming evidence

• category could be noisy but shallow
• some buyers may not yet understand the problem
• blocked sources may hide negative evidence

Recommendation

Start with the productized audit service. Validate with 2 paid pilots before expanding scope.

Open questions

• which niche converts fastest
• which deliverables buyers value most

Sources

• public category pages
• founder discussions
• public pricing and tooling pages

Opportunity Report

Executive summary

Market context

Customer problem signals

Alternatives and competitors

Opportunity hypotheses

Scoring matrix

Risks and disconfirming evidence

Recommendation

Open questions

Sources

Research Brief

Objective

Research question

Scope

Exclusions

Target customer or segment

Geography and timeframe

Success criteria

Key unknowns

Source Log