Writing Statuslines

STARTER_CHARACTER = 📊

Setup

Update the reference docs to get the latest from Anthropic:

bashCopy code
python ~/.claude/skills/writing-statuslines/scripts/update-docs.py

What Status Lines Are

Custom scripts that display contextual information at the bottom of Claude Code's interface. Updated when conversation messages change, at most every 300ms.

Configuration

Add to ~/.claude/settings.json (user-level) or .claude/settings.json (project-level):

jsonCopy code
{
  "statusLine": {
    "type": "command",
    "command": "~/.claude/statusline.sh",
    "padding": 0
  }
}

padding is optional. Set to 0 to let the status line reach the terminal edge.

How It Works

• Claude Code passes session context as JSON via stdin to the script
• First line of stdout becomes the status line text
• ANSI color codes supported
• Script must be executable (chmod +x)
• Only stdout is used (not stderr)

JSON Input Schema

The script receives this via stdin:

jsonCopy code
{
  "hook_event_name": "Status",
  "session_id": "abc123...",
  "transcript_path": "/path/to/transcript.json",
  "cwd": "/current/working/directory",
  "model": {
    "id": "claude-opus-4-1",
    "display_name": "Opus"
  },
  "workspace": {
    "current_dir": "/current/working/directory",
    "project_dir": "/original/project/directory"
  },
  "version": "1.0.80",
  "output_style": {
    "name": "default"
  },
  "cost": {
    "total_cost_usd": 0.01234,
    "total_duration_ms": 45000,
    "total_api_duration_ms": 2300,
    "total_lines_added": 156,
    "total_lines_removed": 23
  },
  "context_window": {
    "total_input_tokens": 15234,
    "total_output_tokens": 4521,
    "context_window_size": 200000,
    "used_percentage": 42.5,
    "remaining_percentage": 57.5,
    "current_usage": {
      "input_tokens": 8500,
      "output_tokens": 1200,
      "cache_creation_input_tokens": 5000,
      "cache_read_input_tokens": 2000
    }
  }
}

context_window.current_usage may be null if no messages have been sent yet.

Key Fields

• model.display_name — short model name ("Opus", "Sonnet")
• workspace.current_dir / workspace.project_dir — may differ when working in subdirectories
• cost.total_cost_usd — cumulative session cost
• context_window.used_percentage / remaining_percentage — pre-calculated, ready to display
• context_window.current_usage — raw token counts from the last API call

Constraints

• Output exactly one line
• Runs every 300ms at most — expensive operations must be cached
• Keep it scannable: glanceable in under a second
• Script must exit cleanly and quickly

Anti-Patterns

• Cramming too much info — pick 3-4 data points max
• Not consuming stdin (script must read it even if it doesn't use all fields)
• Expensive uncached operations (git commands, API calls) on every invocation
• Multiple output lines (only first line is used)
• Forgetting chmod +x
• Writing to stderr instead of stdout

Testing

Test scripts manually with mock JSON:

bashCopy code
echo '{"model":{"display_name":"Sonnet"},"workspace":{"current_dir":"/test"},"cost":{"total_cost_usd":0.05},"context_window":{"used_percentage":42.5}}' | ./statusline.sh

Reference

• references/anthropic-statusline.md - Complete reference with examples in bash, python, and node