deep-research

name

deep-research

description

Use this skill to produce thorough, cited research reports synthesized from multiple web sources. Activate when the user wants to research a topic in depth, do competitive analysis, evaluate a technology, size a market, or run due diligence — anytime an answer requires evidence and source attribution. Triggers include "research", "deep dive", "investigate", "what's the current state of", or "compare X vs Y with sources".

Deep Research

Produce thorough, cited research reports from multiple web sources. Every claim must be backed by a source.

Drift-prone. Web search/scrape tool names, quotas, and result shapes change. Before promising coverage or quoting live source counts, verify which search/fetch tools are actually configured in this environment (re-forge's MCP config in ~/.claude.json, or built-in web search/fetch) and what the current API docs say.

When to activate

• Researching any topic in depth, or competitive/market analysis
• Technology evaluation or due diligence on companies/investors/technologies
• Any question requiring synthesis across multiple sources
• The user says "research", "deep dive", "investigate", or "current state of"

Tooling

Use whatever search and fetch capabilities are available — configured search/scrape MCP tools (e.g. firecrawl, exa) if present, otherwise the built-in web search and web fetch. Confirm availability before promising breadth of coverage.

What to do

1. Understand the goal. Ask 1–2 quick clarifying questions (learning vs. deciding vs. writing? any specific angle or depth?). If the user says "just research it", proceed with reasonable defaults.
2. Plan. Break the topic into 3–5 research sub-questions (e.g. main applications, measured outcomes, regulatory landscape, leading players, market size/trajectory).
3. Execute multi-source search. For each sub-question, search with 2–3 keyword variations, mixing general and news-focused queries. Aim for 15–30 unique sources total. Prioritize academic/official/reputable news > blogs > forums.
4. Deep-read key sources. Fetch full content for the 3–5 most promising URLs — do not rely on search snippets alone.
5. Synthesize and write the report in the structure below.
6. Deliver. Short topics: post the full report in chat. Long reports: post the executive summary + key takeaways and save the full report to a file.

Report structure

# [Topic]: Research Report
*Generated: [date] | Sources: [N] | Confidence: [High/Medium/Low]*

## Executive Summary
[3–5 sentence overview of key findings]

## 1. [Major Theme]
- Key point ([Source Name](url))
- Supporting data ([Source Name](url))

## 2. [Major Theme]
...

## Key Takeaways
- [Actionable insight]

## Sources
1. [Title](url) — one-line summary

## Methodology
Searched [N] queries across web and news. Analyzed [M] sources.
Sub-questions investigated: [list]

Parallel research

For broad topics, parallelize with subagents (one per cluster of sub-questions). Each agent searches, deep-reads, and returns findings; the main session synthesizes the final report.

Hard rules (quality)

• Every claim needs a source — no unsourced assertions.
• Cross-reference: if only one source says it, flag it as unverified.
• Prefer sources from the last 12 months; recency matters.
• Acknowledge gaps explicitly — if a sub-question lacks good info, say "insufficient data found."
• No hallucination. Separate fact from inference; label estimates, projections, and opinions clearly.