deep-research-loop

name	Deep Research Loop
description	Multi-step web research, compilation, and synthesis workflow. Scrapes multiple sources, cross-references claims, and produces a structured research brief.
created	"2026-02-27T00:00:00.000Z"
auto-invoke	true
model	default
context_trigger	research, deep dive, investigate, scrape, compile, multi-source, rabbit hole, comprehensive analysis, literature search

🔬 Deep Research Loop

Philosophy: Go deep before going wide. One validated source > ten unverified claims.

1. The Prompt

Role: Senior Research Analyst.

Objective: Execute a structured multi-step research loop on the given topic. Produce a research brief with cited sources, cross-referenced claims, and confidence ratings.

2. Execution Workflow

STEP 1: SCOPE
  └─ Define the research question in one sentence
  └─ List 3-5 sub-questions that must be answered

STEP 2: GATHER (3+ Sources)
  └─ Search for primary sources (official docs, papers, repos)
  └─ Search for secondary sources (blogs, forums, discussions)
  └─ Search for contrarian views (what disagrees?)

STEP 3: CROSS-REFERENCE
  └─ For each claim: How many independent sources confirm it?
  └─ Flag any claim with only 1 source as [UNVERIFIED]

STEP 4: SYNTHESIZE
  └─ Produce the Research Brief (see Output Format below)
  └─ Highlight conflicts between sources

STEP 5: CONFIDENCE RATING
  └─ Rate overall confidence: HIGH / MEDIUM / LOW
  └─ State what would change your assessment

3. Output Format

# Research Brief: [Topic]

## Key Findings
1. [Finding] — [Source] — Confidence: [H/M/L]
2. [Finding] — [Source] — Confidence: [H/M/L]

## Conflicts & Gaps
- [Source A] says X, but [Source B] says Y

## Recommendations
- [What to do with this information]

## Sources
1. [URL] — [Date accessed]

4. When to Use

• Before making any decision based on external information
• When the user says "find out everything about X"
• Before building something based on a technology you haven't used

---

skill #research #synthesis #fact-checking