How do I start a new planning session?

Simply create a markdown (.md) file containing your requirements and run the command '/deep-plan @path/to/your-spec.md' within Claude Code.

Do I need a specific environment to run Deep Plan?

Yes, Deep Plan requires Python 3.11+ and the 'uv' package manager installed on your system to run the validation, review, and session management scripts.

Does this skill require any external API keys?

While basic planning works within Claude, the 'Deep' functionality requires either a Gemini or OpenAI API key to perform the multi-LLM validation and external review steps.

What does the external LLM review do?

The external review uses your Gemini or OpenAI API keys to provide an independent critique of the generated plan, identifying edge cases, security concerns, or architectural flaws that a single model might miss.

Can I resume a planning session if I get disconnected?

Yes. Deep Plan saves its state in a 'deep_plan_config.json' file within your planning directory, allowing you to resume precisely where you left off by re-invoking the command.

Deep Plan

Name: Deep Plan
Author: piercelamb

bypiercelamb

•

생산성 및 워크플로

Transforms vague requirements into exhaustive, TDD-oriented implementation plans using multi-step research and external LLM reviews.

Deep Plan is a specialized Claude Code skill designed to bridge the gap between high-level ideas and production-ready code. It orchestrates a rigorous multi-stage workflow—including automated codebase and web research, interactive stakeholder interviews, and cross-model validation using Gemini or OpenAI—to generate detailed, sectionized implementation plans. By integrating Test-Driven Development (TDD) principles and splitting complex features into manageable segments, it ensures your development phase is guided by a robust, peer-reviewed technical architecture that minimizes architectural debt.

주요 기능

01Multi-LLM peer review integration with Gemini and OpenAI

028 GitHub stars

03Sectionized planning index for incremental, manageable implementation

04TDD-oriented plan generation with dedicated test stubs

05Automated research subagents for codebase and web analysis

06Interactive stakeholder interview protocol to capture missing requirements

사용 사례

01Developing greenfield projects from high-level or vague specifications

02Architecting complex new features within large, unfamiliar codebases

03Conducting AI-driven peer reviews of technical designs before writing code

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add piercelamb/deep-plan deep-plan

For use in Claude.ai and ChatGPT

Download Skill