How do I apply a custom output style?

Store your style file in ~/.claude/output-styles/ or the project's .claude/output-styles/ directory, then activate it during a session using the /output-style command.

Should I use an output style to enforce coding standards?

No. Output styles are personality suggestions and can be overridden. For actual enforcement of rules or standards, you should use Hooks, which can block operations with exit codes.

When should I use CLAUDE.md instead of an output style?

Use CLAUDE.md for project-specific context such as libraries, conventions, or commands. Use output styles when you want to change 'how' Claude communicates or its specific persona.

Do output styles affect my agents?

No, output styles are orthogonal to agents. Styles only affect the main interactive conversation session, while agents use their own isolated system prompts defined in their agent files.

What is the purpose of the keep-coding-instructions toggle?

When set to false, it removes software engineering assumptions from the system prompt, making it ideal for non-coding tasks like content strategy, research, or general analysis.

Claude Code Output Styles Optimizer

Name: Claude Code Output Styles Optimizer
Author: racurry

byracurry

0•

Learning & Documentation

Optimizes Claude Code interactions by providing architectural guidance on when to use output styles versus agents, hooks, or project context.

This skill provides essential interpretive guidance for Claude Code's output styles, helping users understand the critical distinction between session-level personas and isolated agent prompts. It offers a decision framework to determine whether a workflow benefits from a custom personality, a CLAUDE.md context file, or specialized sub-agents. By clarifying how styles affect the main session without leaking into delegated tasks, this skill ensures developers maintain an efficient, noise-free environment while leveraging specialized learning and explanatory modes.

Key Features

01Offers a strategic decision framework for custom style design and storage

020 GitHub stars

03Provides guidance on using built-in Explanatory and Learning modes for onboarding

04Explains the keep-coding-instructions toggle for non-engineering domains like technical writing

05Identifies common pitfalls such as using styles for enforcement instead of hooks

06Clarifies the architectural split between session-wide output styles and isolated agents

Use Cases

01Onboarding to an unfamiliar legacy codebase using Explanatory mode insights

02Transitioning Claude Code into a non-engineering persona for research and documentation analysis

03Determining the most efficient way to implement project-wide coding standards versus persona shifts

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add racurry/neat-little-package output-styles

For use in Claude.ai and ChatGPT

Download Skill