Back to Skills

ralph-plan

Interactive planning assistant that helps create focused, well-structured ralph-loop commands through collaborative conversation

26,710stars2,533forksUpdated 7/30/2026

Security Assessment

Safe(95/100)
Security Score95/100

About ralph-plan

This skill is an interactive planning assistant that helps users author well-structured "ralph-loop" commands through collaborative conversation. It solves the problem of turning a vague goal into a focused, actionable, and verifiable plan by guiding the user step by step rather than producing a full command in one shot.

The skill defines a fixed ralph command structure with background, setup, tasks, and testing sections plus a completion promise, and it drives a five-step planning process: understand the goal, define the background and persona, plan setup steps, break the goal into concrete numbered tasks ordered by dependency, and define testing and success criteria. It emphasizes being inquisitive (probing for edge cases and assumptions), identifying gaps, and proactively researching the codebase to fill knowledge gaps so tasks can reference specific file paths, function names, and existing patterns. It includes an example conversation flow and an output format note advising against double quotes and backticks in the final command to avoid copy/paste issues. The skill is pinned to a specific Claude model in its front matter.

The target users are developers working within the Mastra project (or similar setups) who use the ralph-loop workflow to drive autonomous or semi-autonomous coding tasks and want help scoping and structuring those runs. Use cases include planning a new feature, breaking a large task into a focused command, and ensuring a plan includes the research and testing steps needed to be executed reliably. It is a planning and prompt-authoring helper with no execution side effects of its own.

FAQ

What does this skill produce?

A structured ralph-loop command with background, setup, tasks, and testing sections, ending with a COMPLETE promise, built collaboratively with the user.

How does the planning process work?

Five steps: understand the goal, define background and persona, plan setup, break down numbered tasks ordered by dependency, and define testing and success criteria, refined iteratively through conversation.

Does it explore the codebase?

Yes. It is designed to proactively research existing code to suggest specific file paths, function names, and patterns rather than relying only on what the user describes.

Are there formatting constraints on the output?

It advises avoiding double quotes and backticks in the final command, using single quotes or rephrasing, so the command survives copy and paste without breaking.

Does the skill itself run code or make changes?

No. It is a conversational planning assistant that helps author a command; execution of that command happens separately.

Install ralph-plan

Download and extract the skill files to your .claude/skills/ directory.

Quick Setup:

  1. Copy the skill folder to .claude/skills/
  2. Claude will automatically detect and use the skill