What problem does it solve?

Engineers and decision-makers need trustworthy multi-source research within Claude Code, but built-in WebSearch only returns titles/URLs and has no source quality evaluation. Existing research skills either delegate to external APIs (Gemini) or lack source verification.

Deep Research runs a 5-phase parallel pipeline entirely within Claude Code — no API keys, no external dependencies — with hard quality gates that block output unless every claim cites a classified source.

Who uses this workflow?

Software engineers evaluating technical decisions (e.g., "should we migrate from Redis to Valkey?"), teams doing competitive analysis, and anyone who needs a well-sourced answer in minutes rather than hours.

What makes it different?

• 4-tier source classification — every source labeled as Primary, Curated Secondary, Community, or Unverifiable
• 5 hard quality gates — no fabricated URLs, no uncited findings, failed angles disclosed
• Confidence labels — Established / Likely / Emerging / Contested / Speculative
• 3 scope modes — quick (~2 min), standard (~5 min), thorough (~10 min)
• Codebase-aware — spawns Explore agents for project-related questions
• No external API keys required — uses built-in WebSearch + WebFetch

Usage example

/deep-research what are the tradeoffs between SQLite and PostgreSQL for new web apps?
/deep-research quick what is WebTransport?
/deep-research thorough should we migrate from Redis to Valkey?

Attribution

Created by @robertnowell. MIT licensed.