Output Styles & Status Lines

What Output Styles Do

Output styles replace parts of Claude Code's default system prompt with your own instructions. The model and tool capabilities stay identical — only the communication protocol changes. Built-in styles include Default, Explanatory, and Learning. Custom styles can produce bullets, YAML, HTML, or any format you define.

Creating a Custom Output Style

Output styles are Markdown files with YAML frontmatter. User-scoped go in ~/.claude/output-styles/, project-scoped in .claude/output-styles/. Use /output-style to switch active style or /output-style:new to create one with help from Claude.

---
description: Concise bullet-point responses with direct communication
---

# Communication Style
- Use bullet points for all responses
- Be direct and to the point
- No unnecessary language or filler

# Response Format
- Lead with the answer
- Use nested bullets for hierarchy
- Skip preamble and transitions

# Tone
- Professional and concise
- No hedging or qualifications

Why YAML as an Output Style

YAML is more than formatting — it is implicit semantic structure. Indentation expresses scope, key ordering encodes priority, nesting maps dependencies. Asking Claude to respond in YAML changes how it reasons: linear narratives become hierarchical models with relationships, scope, and priority encoded structurally.

Task: "Fix authentication bug"
Priority: "Urgent high"

approach:
  investigate:
    auth_flow:
      - login_endpoint
      - token_validation
      - session_middleware

fix:
  files_to_modify:
    auth.js:
      scope: "token refresh logic"
      risk: "low"
    middleware.js:
      scope: "validation checks"
      risk: "urgent"

verify:
  tests: ["auth_integration", "token_expiry"]

Embedding Workflows in Output Styles

Output styles can encode workflow logic, not just formatting. An HTML output style can instruct Claude to save the generated HTML to a .html file and open it in the default browser — eliminating the need to specify these steps in every prompt.

---
description: Format responses as retro HTML pages with ASCII art styling
---

Format all responses as complete HTML5 documents with retro ASCII art styling.

# Workflow
- Save the HTML file after writing it with a descriptive name ending in .html
- OPEN the generated file in the default web browser

# HTML Structure
- Always provide complete HTML5 with DOCTYPE, head, body
- Use semantic elements (header, main, section, article, footer)

# Styling
- Embed CSS in <style> tags in the HTML head
- Monospace fonts (Courier New, Monaco, Consolas)
- Dark backgrounds with green-on-black or amber-on-black schemes

Status Lines

Status lines display persistent info at the bottom of the terminal. They run any executable script — bash, Python, Node — that receives Claude Code's session JSON via stdin and prints the line to stdout. Configure via /statusline slash command or by editing ~/.claude/settings.json directly.

{
  "model": "sonnet",
  "statusLine": {
    "type": "command",
    "command": "uv run ~/.claude/statusline.py"
  }
}

The script receives session data — model, context usage, costs, git info, output style, and transcript_path — as JSON on stdin. Useful displays: active output style, last user prompt, current git branch, token usage percentage. ANSI escape codes work for color and bold formatting.