output-styles-guide
Adjusts Claude Code system prompts to change behavior for non-coding tasks like teaching or learning.
Install
mkdir -p .claude/skills/output-styles-guide && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/6688" && unzip -o skill.zip -d .claude/skills/output-styles-guide && rm skill.zipInstalls to .claude/skills/output-styles-guide
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Adapt Claude Code for different use cases beyond software engineering by customizing system prompts for teaching, learning, analysis, or domain-specific workflows.Key capabilities
- →Modify system prompts to adapt Claude Code's behavior
- →Exclude efficiency-focused coding instructions when not needed
- →Inject custom instructions tailored to a specific role or workflow
- →Persist output style settings at project or user level
- →Create new output styles using a guided command or manual file creation
- →Switch between built-in styles like Default, Explanatory, and Learning
How it works
Output styles modify Claude Code's system prompt to change its behavior for specific use cases, allowing it to prioritize different types of interactions or information.
Inputs & outputs
When to use output-styles-guide
- →Switch to explanatory mode for code review
- →Set up collaborative learning workflow
- →Customize output for documentation
- →Configure project-specific AI behaviors
About this skill
Output Styles Guide
Output styles adapt Claude Code's behavior for use cases beyond software engineering by modifying the system prompt, enabling specialized workflows while retaining core capabilities (file I/O, script execution, TODO tracking).
What Are Output Styles?
Output styles are customized system prompts that replace or supplement Claude Code's default software engineering focus. They:
- Exclude efficiency-focused coding instructions when not needed.
- Inject custom instructions tailored to a specific role or workflow.
- Persist at project level (
.claude/settings.local.json) or user level (~/.claude/output-styles). - Preserve tool access (Bash, file editing, TODO management).
Built-In Output Styles
Default
Standard Claude Code behavior optimized for software engineering: concise output, code verification, efficient task completion.
Explanatory
Adds "Insights" sections between tasks to explain implementation choices and codebase patterns. Ideal for understanding complex code or teaching.
Learning
Collaborative, learn-by-doing mode. Shares insights and requests your contribution on small code sections via TODO(human) markers. Best for skill-building or onboarding.
Changing Your Output Style
Interactive menu:
/output-style
# or access via /config
Direct command:
/output-style explanatory
/output-style default
/output-style learning
Changes apply at project level and save to .claude/settings.local.json.
Creating Custom Output Styles
Quick start (guided):
/output-style:new I want an output style that [describes your use case]
Claude creates and saves a template; you refine it.
Manual creation:
Create a markdown file at ~/.claude/output-styles/<name>.md (user-level, shared across projects) or .claude/output-styles/<name>.md (project-level only).
Structure:
---
name: My Custom Style
description: Brief description shown in /output-style menu
---
# Custom Style Instructions
You are an interactive CLI tool. [Your instructions here...]
## Specific Behaviors
[Define how the assistant behaves...]
Example: Research Assistant Style
---
name: Research Assistant
description: Focused, depth-first analysis with citations and hypothesis tracking.
---
# Research Assistant Mode
You are a research partner specializing in deep investigation and synthesis.
## Specific Behaviors
- Request sources and cite evidence when making claims.
- Track open hypotheses explicitly.
- Summarize findings in bullet-point format with confidence levels.
- Flag uncertainty and propose next investigation steps.
Best practices for custom styles:
- Be specific: "summarize in 3 bullets", "include citations", "ask for feedback".
- Retain tool flexibility: don't disable essential capabilities unless necessary.
- Test with a few tasks to verify behavior before distributing.
Common Use Cases
| Use Case | Style | Benefit |
|---|---|---|
| Learning codebase | Explanatory | Understand why code is structured this way |
| Onboarding engineers | Learning | Active participation, hands-on skill building |
| Research/analysis | Custom | Depth-first investigation, hypothesis tracking |
| Technical writing | Custom | Structured outlines, examples, glossary generation |
| Product/UX work | Custom | Personas, user flows, journey mapping focus |
Output Styles vs. Related Features
| Feature | Purpose | Scope |
|---|---|---|
| Output Styles | Persistent system prompt modification | Affects all main agent interactions |
| CLAUDE.md | Project-level instructions added after system prompt | Supplements default behavior; doesn't replace it |
| --append-system-prompt | Runtime system prompt additions | One-time append per session |
| Agents | Task-specific execution with custom tools/models | Single-purpose delegation; doesn't affect main loop |
| Custom Slash Commands | Stored user prompts (input templates) | Shorthand for repeated requests |
Key distinction: Styles replace core system instructions; others add to them.
Tips & Troubleshooting
- Not persisting? Verify save location:
.claude/settings.local.jsonfor project,~/.claude/output-styles/for user-level styles. - Lost formatting? Keep custom style descriptions under 100 chars for menu readability.
- Want to share? Save custom styles at project level (
.claude/output-styles/) and commit to Git. - Reverting? Run
/output-style defaultor delete from.claude/settings.local.json. - Stacking instructions? Use CLAUDE.md alongside styles to add project-specific rules to your custom style.
Quick Reference
| Action | Command |
|---|---|
| View available styles | /output-style |
| Switch directly | /output-style [style-name] |
| Create custom | /output-style:new [description] |
| Open config | /config |
| Access settings | .claude/settings.local.json (project) or ~/.claude/output-styles/ (user) |
Limitations
- →Custom style descriptions should be under 100 characters for menu readability
- →Output styles affect all main agent interactions
- →Does not disable essential capabilities unless necessary
How it compares
This approach replaces core system instructions, unlike other features that only add to the system prompt or provide one-time additions.
Compared to similar skills
output-styles-guide side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| output-styles-guide (this skill) | 1 | 10mo | Review | Beginner |
| system-prompt-writer | 6 | 9mo | No flags | Intermediate |
| skills | 6 | 6mo | No flags | Intermediate |
| review-claudemd | 1 | 6mo | Review | Advanced |
Try saying
Example prompts that trigger this skill in your AI assistant.
More by CaptainCrouton89
View all by CaptainCrouton89 →You might also like
system-prompt-writer
aws-samples
This skill should be used when writing or improving system prompts for AI agents, providing expert guidance based on Anthropic's context engineering principles.
skills
atopile
How to write and maintain `.claude/skills/*/SKILL.md` files: source-of-truth-first process, verification steps, and conventions.
review-claudemd
ykdojo
Review recent conversations to find improvements for CLAUDE.md files.
writing-documentation-for-llms
CaptainCrouton89
Teaches how to write effective documentation and instructions for LLMs by assuming competence, using progressive disclosure, prioritizing examples over explanations, and building feedback loops into workflows.
intelligent-prompt-generator
huangserva
智能提示词生成器 v2.0 - 支持人像/跨domain/设计三种模式,语义理解、常识推理、一致性检查
skills-guide
CaptainCrouton89
Defines required structure, frontmatter format, and best practices for SKILL.md files. Use BEFORE creating or editing any skill - this is the spec to follow, not optional reference.