Deep Research Skill
by Community
Community-maintained deep-research skill for Claude. Plan, search, synthesize, cite. The discovery loop, packaged.
Skills
Deep Research Skill
Added 17 May 2026
Overview
The Deep Research skill encodes a reliable plan-search-synthesize loop into a Claude skill. It scaffolds the agent's research session: scope, queries, source quality checks, structured notes, then a final synthesis with citations. The right pattern when you want consistent research output across many runs.
Best for
Best for
Teams running research workflows where structure beats raw token spend
Use cases
- Run a market-scan research pass with consistent structure
- Produce a citation-backed briefing on a new topic
- Standardise research output across a team of analysts
- Bridge research outputs into downstream report pipelines
Notes
Why it matters
Research is one of the highest-value agent workflows and one of the easiest to do badly. A skill that locks in a real research loop lifts every research session above the random-chat baseline.
How teams use it in production
Pair with a search MCP server (Brave, Tavily, or a private corpus). Let the skill run the structured loop. Review the synthesis, not every step in the middle.
What to watch
Hosted research products (Perplexity, ChatGPT Pro, Anthropic Research) and local research skills are converging on the same structure. The locally-runnable skill keeps the data inside your boundary.
Pros
- Structured output across runs, not free-form summaries
- Forces citations and source quality checks
- Pairs cleanly with web search MCP servers
- Easy to fork for org-specific research checklists
Cons
- Quality bounded by the search tool the agent has access to
- Deep niche topics still need expert review
- Citations need a verification pass on high-stakes work
Pairs with
Other entries in the index that connect to this one. Click through to see the chain.