How does this improve the quality of AI-generated code?

By forcing the AI to build a mental model first, it maintains better context, adheres more closely to existing patterns, and considers edge cases often missed when coding on the fly.

Can I skip phases for simple bug fixes?

Yes, the skill includes a complexity matrix that suggests skipping the explore and plan phases for trivial tasks while mandating them for complex features.

How does this skill prevent premature coding?

It enforces a mandatory 'Explore' phase where the AI must read and understand existing files and an 'Approve' phase that requires user validation of a plan before any code is generated.

Why is the 'Approve' phase necessary?

The Approve phase creates a vital human-in-the-loop checkpoint, allowing you to catch logic errors or architectural misalignments before time is spent on implementation.

When should I use the Explore-Plan-Code skill?

Use this skill for any non-trivial task, such as new feature development, complex refactors, or architectural changes where understanding system dependencies is critical.

Explore-Plan-Code Workflow

Name: Explore-Plan-Code Workflow
Author: rahulsub

byrahulsub

•

生産性とワークフロー

Enforces a structured four-phase development methodology to prevent premature coding and ensure architectural consistency during complex tasks.

The Explore-Plan-Code skill optimizes Claude's development output by implementing a rigorous research-first workflow designed for non-trivial software engineering tasks. By separating the process into distinct Explore, Plan, Approve, and Code phases, the skill ensures the AI builds a comprehensive mental model of the codebase before proposing changes. This deliberate approach minimizes technical debt, prevents false starts, and allows for human-in-the-loop verification of architectural decisions before any implementation begins, leading to significantly more robust and maintainable code.

主な機能

01Detailed planning prompts utilizing extended reasoning for architectural design

02Mandatory approval gates to ensure human alignment on technical strategies

03Pre-defined testing and rollback strategies included in every plan

04Complexity-based scaling for different task types from trivial to architectural

051 GitHub stars

06Context-aware exploration that builds mental models before implementation

ユースケース

01Major architectural changes requiring deep reasoning and risk assessment

02Implementing complex features that span multiple services or directories

03Refactoring legacy code where dependencies and data flows are unclear

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add rahulsub/code-practices-plugin explore-plan-code

For use in Claude.ai and ChatGPT

Download Skill