Deep Research (claude-skills)

Deep Research (claude-skills) is a community plugin from Alireza Rezvani's claude-skills collection: the heavyweight research workflow for questions where a wrong answer is expensive. It gives Claude a nine-phase method, a research agent and a /cs-deep-research command, and saves each investigation as a folder you can come back to. The method was contributed by Socialpranker from their MIT-licensed Deepdive project, and it's credited in the plugin. What it does - Reframes your question around the decision behind it, with two to four falsifiable hypotheses - Writes plan.md: scope, sourcing strategy, opposition queries, a risk register and stop criteria - Checks which search channels and API keys you have, and falls back to plain web pages - Sends search sub-agents out in parallel and saves each source to its own file, with verbatim quotes - Scores every source for credibility, recency and bias, and backs each thesis with at least three independent sources of different types, or marks it "insufficient evidence" - Runs an adversarial pass with steel-manned counter-arguments, then a light citation check - Writes refresh_targets.md, so a later "update <slug>" run produces a delta instead of starting over What you get A folder with plan.md, sources.csv, one file per source, findings, the final report and refresh_targets.md. When to use it - Strategy work, a business plan, or the groundwork for a report or article - Comparing several products, methods, markets or institutions with reasoning you can defend - Validating a hypothesis or a decision against outside data - Mapping how a field works It isn't meant for quick fact checks or fast overviews; the same collection's research router covers those. How to install it In Claude Code, run /plugin marketplace add alirezarezvani/claude-skills, then /plugin install deep-research@claude-code-skills. Run it with /cs-deep-research and your question, or ask for "deep research on" a topic. Good to know - It's a method, not a data source: it searches with your agent's own web search and fetch tools, plus any APIs you already have keys for. - It trades tokens for rigor: its agent is set to Opus and runs several search sub-agents at once. - The full source catalog (search channels, statistical sources, APIs and report blocks) lives in Socialpranker's upstream project, which the skill points to. - It's a community plugin, MIT-licensed.

• “Deep research on whether we should move our product to usage-based pricing”

  Copy

• “Compare Postgres logical replication with CDC tools and give me defensible reasoning”

  Copy

• “Validate this hypothesis with external data: mid-size retailers are cutting loyalty programs”

  Copy

• “Map the landscape of open-source vector databases”

  Copy

• “Research this rigorously: is now a good time to enter the EU e-bike market?”

  Copy

• “Update my vector-databases research with what's changed”