deepthinklite

Automate deep research with local-first artifact generation and configurable time budgets.

Updated Feb 9, 2026
One-click install
npx skills add https://github.com/VirajSanghvi1/deepthinklite-skill --skill deepthinklite
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: deepthinklite
Source: https://github.com/VirajSanghvi1/deepthinklite-skill/tree/main/skills/deepthinklite
Command: npx skills add https://github.com/VirajSanghvi1/deepthinklite-skill --skill deepthinklite

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

DeepthinkLite addresses the challenge of preserving research work and producing structured, reusable outputs by enabling a local-first deep-research workflow that saves progress as artifacts.

Core Features & Use Cases

  • Two durable artifacts: questions.md (investigation map) and response.md (final answer) that you can diff and version-control.
  • Local-first workflow: keeps all work in your repository and avoids losing chat history.
  • Time-budgeted research: enforces a configurable minimum and maximum time to ensure thoroughness and guardrails.
  • Safety and permissions: explicit prompts and untrusted-source handling to avoid prompt-injection and data leakage.
  • Modes of source input: supports raw snippets or summary-only inputs with user permission prompts.

Quick Start

Invoke the DeepthinkLite tool with your research question to create a run directory containing questions.md and response.md.

Frequently Asked Questions about deepthinklite

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I save deep research progress as reusable artifacts in my local repository?

Deep research progress is saved as reusable artifacts by generating questions.md and response.md directly in your local repository. This local-first workflow ensures your investigation map and final answers are durable and version-controlled.

What is a time-budgeted research workflow and how does it enforce thoroughness?

A time-budgeted research workflow enforces configurable minimum and maximum time limits to ensure deep research thoroughness. These guardrails prevent both premature termination and excessive processing time on a single investigation.

How do I prevent prompt injection when researching untrusted sources?

Prevent prompt injection during deep research by using a workflow with explicit user permission prompts and untrusted-source handling. It supports raw or summary-only source input modes to actively mitigate data leakage risks.

Does deep research work with local-first workflows without losing chat history?

Yes, deep research works with local-first workflows by keeping all progress and generated artifacts in your local repository. This approach avoids losing chat history and ensures your research outputs remain fully version-controlled.

What is the best way to structure deep research outputs for version control?

The best way to structure deep research outputs for version control is a two-artifact model: questions.md for the investigation map and response.md for the final answer. These markdown files are easily diffable within any repository.

Can I use summary-only source inputs for deep research with explicit permission prompts?

Yes, you can use summary-only source inputs for deep research. The workflow supports both raw snippets and summary-only modes, explicitly requiring user permission prompts before processing untrusted sources to ensure safety.