abhiroopb/skills/web-research

Web Research

Search the web, pull in relevant content, and synthesize findings into a structured research brief. Designed to complement PRD-writing and spec-creation workflows by providing external context — competitor analysis, market trends, industry benchmarks, best practices, and technical reference material.

When to Use

• Researching competitors or comparable products before writing a PRD
• Gathering market context, industry trends, or benchmarks for a feature area
• Finding best practices or design patterns for a UX/product decision
• Looking up technical concepts, API docs, or platform capabilities
• Answering "how do others solve this?" or "what's the state of the art?"
• Augmenting internal research (Slack, Snowflake, Glean) with external perspective

Workflow

1. Clarify the Research Goal

Identify from the user's request:

• Topic: What are we researching?
• Purpose: Why? (e.g., PRD background, competitive analysis, technical feasibility)
• Scope: Broad landscape scan or deep dive on a specific question?
• Output format: Research brief, comparison table, bullet summary, or raw findings?

If the purpose is unclear, ask one focused question. Default to a structured research brief.

2. Generate Search Queries

From the topic, generate 3–7 diverse search queries that cover:

• Direct queries: The topic itself (e.g., "competitor checkout flow comparison")
• Comparative queries: "best [category] solutions 2025", "[competitor] vs [competitor]"
• Technical queries: Implementation details, API docs, platform specifics
• Trend queries: "state of [industry] 2025", "[topic] market trends"

Use web_search with each query. Set max_results to 5–8 per query.

3. Read and Extract

For the most promising results (top 5–10 across all queries):

• Use read_web_page with a focused objective to extract only relevant content
• Skip paywalled, login-gated, or low-quality sources
• Prioritize: official docs, reputable publications, industry reports, product pages

4. Synthesize Findings

Compile findings into a structured research brief:

# Web Research: <Topic>

**Research goal**: <one-line purpose>
**Date**: <today's date>
**Queries used**: <list of search queries>

## Key Findings

### <Finding Category 1> (e.g., Competitor Landscape)
- **[Source Name](URL)**: <key insight or data point>
- **[Source Name](URL)**: <key insight or data point>

### <Finding Category 2> (e.g., Best Practices)
- **[Source Name](URL)**: <key insight or data point>

### <Finding Category 3> (e.g., Market Data)
- **[Source Name](URL)**: <key insight or data point>

## Summary

<3–5 sentence synthesis of the most important takeaways>

## Relevance to <Purpose>

<How these findings inform the PRD/spec/decision — specific recommendations>

## Sources

| # | Source | URL | Relevance |
|---|--------|-----|-----------|
| 1 | [Name] | [URL] | [Why it matters] |

5. Present and Offer Next Steps

After presenting the brief, offer:

• "Want me to deep-dive on any of these findings?"
• "Want me to use these findings to start a PRD?" (→ load writing-requirements-docs)
• "Want me to save this research to a file?"

If the user wants to save, write to research/<topic-kebab-case>/WEB-RESEARCH.md in the current working directory.

Integration with Other Skills

This skill is designed to chain with:

• writing-requirements-docs: Provide external context for Problem Statement (§1), competitive landscape for Non-Goals (§3), and market data for Goals (§2)
• spec-creator: Supply technical research and best practices
• feedback-searcher: Complement internal seller feedback with external market perspective
• prototype-builder: Research UI patterns and competitor implementations before building

When called as part of a PRD workflow, skip the standalone brief format and return findings as inline-ready content (quoted with source links) that the calling skill can weave directly into the document.

Quality Standards

• Every claim must have a source link. No unsourced assertions.
• Prefer recent sources. Prioritize content from the last 12 months.
• Distinguish facts from opinions. Label analyst opinions, projections, and estimates clearly.
• Be concise. Summarize; don't dump raw page content.
• Note gaps. If a question couldn't be answered from web sources, say so explicitly.

Anti-patterns

• Don't return raw search results without reading and synthesizing
• Don't include sources that are paywalled or inaccessible
• Don't present speculation as fact
• Don't over-research — aim for 5–10 high-quality sources, not 50 mediocre ones
• Don't skip source attribution — every finding needs a clickable URL