This project provides a structured, human-in-the-loop deep research skill designed for AI models like Claude Code, OpenCode, and Codex. It addresses the need for systematic information gathering and analysis across various domains, including academic, technical, market research, and due diligence, offering users precise control over the research process and delivering organized, actionable reports.

How It Works

The skill employs a two-phase research workflow: initial outline generation and subsequent deep investigation. Users guide the process, defining research topics and refining outlines before initiating an automated deep dive. This deep research phase leverages parallel agents to efficiently gather comprehensive data on each research item, ensuring thoroughness and speed. The human-in-the-loop design allows for iterative refinement, making it suitable for complex research tasks requiring accuracy and specific focus.

Quick Start & Requirements

Installation involves cloning the repository (git clone https://github.com/Weizhena/deep-research-skills.git) and navigating into the directory. A primary Python dependency is pip install pyyaml. Specific setup varies by model:

• Claude Code: Copy skill files to ~/.claude/skills/ and agent files to ~/.claude/agents/.
• OpenCode: Copy skill files, enable web search via export OPENCODE_ENABLE_EXA=1 (persistent via ~/.bashrc), and copy agent files to ~/.config/opencode/agents/.
• Codex: Copy skill and agent files to ~/.codex/ directories, or use the provided bash scripts/install-codex.sh for automatic installation.

Highlighted Details

• Supports structured research for academic papers, technical evaluations, market analysis, and due diligence.
• Commands include /research (outline generation), /research-add-items, /research-add-fields, /research-deep (deep investigation), and /research-report (markdown report generation).
• The deep research phase utilizes parallel agents for efficient data collection.
• Inspired by RhinoInsight, focusing on control mechanisms for model behavior.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or roadmap are provided in the README.

Licensing & Compatibility

The project is released under the MIT License, which is permissive and generally compatible with commercial use and closed-source applications.

Limitations & Caveats

The README does not explicitly detail any limitations, alpha status, or known bugs. The effectiveness of the deep research phase may depend on the underlying AI model's capabilities and the quality of web search results.