How does the Deep Research skill handle large-scale tasks?

It decomposes the main goal into smaller sub-tasks and uses codex exec to run them in parallel, significantly reducing the time required for comprehensive data gathering.

Where are the research results stored?

All findings, logs, and the final refined report are saved in a structured .research/ directory within your workspace, ensuring files are organized and persistent.

What tools does this skill use for internet access?

The skill prioritizes installed Claude skills first, followed by MCP servers like Firecrawl and Tavily, and only uses direct shell commands as a last resort.

Does this skill work with any LLM model?

It is designed to work with your default Claude Code settings and avoids overriding model configurations unless explicitly requested by the user.

Is the research process interactive?

No, the sub-processes are designed to be non-interactive and run in the background, outputting logs and files for the main agent to aggregate without constant user prompting.

Deep Research Orchestrator

Name: Deep Research Orchestrator
Author: micsapp

bymicsapp

0•

Web Scraping y Recopilación de Datos

Automates multi-agent deep research workflows by orchestrating parallel sub-tasks to generate comprehensive, data-driven reports.

This skill provides a sophisticated framework for conducting exhaustive research by breaking high-level objectives into parallelizable sub-goals executed in isolated sandboxes. It coordinates the entire research lifecycle—from initial target decomposition and autonomous web scraping using integrated skills or MCP tools like Firecrawl, to algorithmic aggregation and iterative chapter-by-chapter refinement. By prioritizing structured file-based delivery over chat summaries, it ensures that users receive production-ready, evidence-backed reports suitable for competitive analysis, technical due diligence, or complex documentation projects.

Características Principales

010 GitHub stars

02Parallel sub-task orchestration using codex exec in secure sandboxes

03Prioritized networking using specialized skills and MCP tools like Firecrawl and Tavily

04Autonomous goal decomposition into manageable, multi-agent workstreams

05Automated script-based data aggregation with cross-referenced evidence

06Iterative chapter-by-chapter refinement to ensure high-quality, structured document output

Casos de Uso

01Long-form investigative writing requiring multi-source evidence synthesis

02Systematic data collection and verification for technical documentation or datasets

03Comprehensive competitor and industry market trend analysis

Características Principales

010 GitHub stars

02Parallel sub-task orchestration using codex exec in secure sandboxes

03Prioritized networking using specialized skills and MCP tools like Firecrawl and Tavily

04Autonomous goal decomposition into manageable, multi-agent workstreams

05Automated script-based data aggregation with cross-referenced evidence

06Iterative chapter-by-chapter refinement to ensure high-quality, structured document output

Casos de Uso

01Long-form investigative writing requiring multi-source evidence synthesis

02Systematic data collection and verification for technical documentation or datasets

03Comprehensive competitor and industry market trend analysis