Complete List of Caveman Slash Commands in Claude Code

All Caveman slash commands are invoked with a leading / and defined in the opencode plugin, supporting compression levels, classical Chinese modes, file operations, and utility functions.

The Caveman plugin for Claude Code provides a terse, token-efficient communication mode inspired by primitive speech patterns. According to the JuliusBrussee/caveman source code, all functionality is exposed through slash commands that compress or transform text output. Below is the complete reference of available commands.

Core Caveman Mode Commands

These commands control the active compression level. The parser lives in src/plugins/opencode/commands/caveman.md.

• /caveman — Activates Caveman mode at full strength (default). Drops non-essential tokens aggressively.
• /caveman lite — Light compression, reducing output tokens by approximately 30%.
• /caveman ultra — Maximum compression; removes as many tokens as possible while preserving meaning.
• /caveman off — Deactivates Caveman mode and returns to standard language output.

Wen-Yan (Classical Chinese) Variants

Caveman supports Wen-Yan compression using classical Chinese grammar and vocabulary. Three strength levels mirror the standard modes:

• /caveman wenyan-lite — Light classical Chinese compression.
• /caveman wenyan-full — Standard classical Chinese compression.
• /caveman wenyan-ultra — Maximum classical Chinese compression.

These variants are enumerated in src/plugins/opencode/commands/caveman-help.md alongside the core commands.

Utility and Workflow Commands

Beyond mode switching, Caveman provides specialized slash commands for common development tasks. Each command has its own implementation file in the src/plugins/opencode/commands/ directory.

/caveman-commit

Generates a terse commit message following Caveman style conventions.

Implemented in src/plugins/opencode/commands/caveman-commit.md, this command analyzes staged changes and produces a one-line summary using compressed, high-information-density language.

# Example usage

/caveman-commit

/caveman-review

Runs a one-line code review with severity indicators.

The caveman-review command (source: src/plugins/opencode/commands/caveman-review.md) scans the current diff and emits findings using emoji-based severity markers (e.g., 🔴 for critical, 🟡 for warnings, 🟢 for suggestions).

# Example usage

/caveman-review

/caveman-compress

Compresses a Markdown file by dropping filler text.

This command accepts a filename argument and applies the currently active Caveman compression level. The implementation in src/plugins/opencode/commands/caveman-compress.md handles file I/O and token pruning.

# Compress notes.md using active Caveman level

/caveman-compress notes.md

/caveman-stats

Displays lifetime token-savings statistics.

Sourced from src/plugins/opencode/commands/caveman-stats.md, this command reads the persistent history log and reports:

• Total tokens saved across all sessions
• Number of Caveman sessions
• Average compression ratio

# Example usage

/caveman-stats

Quick Reference: All Slash Commands

Slash Command	Purpose	Source File
/caveman	Activate full Caveman mode	caveman.md
/caveman lite	~30% token reduction	caveman.md
/caveman ultra	Maximum token removal	caveman.md
/caveman wenyan-lite	Light classical Chinese	caveman-help.md
/caveman wenyan-full	Standard classical Chinese	caveman-help.md
/caveman wenyan-ultra	Maximum classical Chinese	caveman-help.md
/caveman off	Return to normal language	caveman.md
/caveman-commit	Terse commit message	caveman-commit.md
/caveman-review	One-line code review	caveman-review.md
/caveman-compress <file>	File compression	caveman-compress.md
/caveman-stats	Token savings report	caveman-stats.md

Natural Language Alternatives

The parser in caveman.md also recognizes natural language equivalents. While the slash syntax is canonical in Claude Code, you may use phrases like "turn on caveman", "stop caveman", or "normal mode" to trigger the same behaviors.

Summary

• Nine slash commands control Caveman functionality in Claude Code, ranging from /caveman (mode activation) to /caveman-stats (analytics).
• Three compression tiers exist for both standard and Wen-Yan modes: lite, full (default), and ultra.
• Four utility commands handle commit messages, code reviews, file compression, and statistics—each implemented in separate files under src/plugins/opencode/commands/.

Frequently Asked Questions

How do I deactivate Caveman mode in Claude Code?

Use /caveman off to exit Caveman mode and restore normal language output. The command parser in src/plugins/opencode/commands/caveman.md handles this state transition and clears any active compression flags.

What is the difference between /caveman ultra and /caveman wenyan-ultra?

Both apply maximum compression, but /caveman ultra uses abbreviated English/Universal grammar while /caveman wenyan-ultra compresses into Classical Chinese (Wen-Yan) constructs. The Wen-Yan variants are documented in src/plugins/opencode/commands/caveman-help.md as distinct command families.

Can I compress existing files with Caveman?

Yes. The /caveman-compress <file> command reads a Markdown file and writes a compressed version using your currently selected Caveman level. The implementation in src/plugins/opencode/commands/caveman-compress.md preserves structural elements (headers, lists, code blocks) while removing filler words and redundant phrases.

Where are Caveman token statistics stored?

The /caveman-stats command reads from a persistent history log maintained by the plugin. As implemented in src/plugins/opencode/commands/caveman-stats.md, it aggregates total tokens saved, session counts, and compression ratios across all historical Caveman usage sessions.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}