Ultra-compressed communication mode for Gemini CLI. Cuts ~75% of tokens while keeping full technical accuracy by making the AI speak like a caveman.

  gemini extensions link /home/christopher/Projects/gemini-caveman

• Token Efficiency: Saves tokens and context window space by removing fluff, pleasantries, and filler words.
• Multiple Intensities: Choose from lite (professional but tight) to ultra (maximum abbreviation) or classical Chinese (wenyan).
• Auto-Clarity: Automatically drops caveman mode for security warnings or complex irreversible actions to ensure safety.
• Custom Commands: Built-in commands for ultra-compressed commit messages (/caveman/commit) and code reviews (/caveman/review).
• Smart Context: Automatically tracks your preferred mode across sessions using hooks.

Clone this repository and add it to your Gemini CLI workspace or global extensions directory. The CLI will automatically load the commands and hooks.

Toggle or adjust the intensity of the caveman communication mode:

/caveman             # Activate default (full) caveman mode
/caveman lite        # No filler. Keep articles + full sentences
/caveman ultra       # Maximum abbreviation, strip conjunctions, use arrows
/caveman wenyan-full # Maximum classical Chinese terseness
/caveman off         # Return to normal verbose communication

• /caveman/commit Generate an ultra-compressed commit message in Conventional Commits format for your staged changes. No fluff, no "This commit does X".

• /caveman/review Generate terse, actionable code review comments. One line per finding. Format: L<line>: <severity>: <problem>. <fix>.

Standard AI:

"Sure! I'd be happy to help you with that. The issue you're experiencing is likely caused by the fact that the token expiry check uses < instead of <=."

Caveman (full):

"Bug in auth middleware. Token expiry check use < not <=. Fix:"

Standard Review:

"It looks like this function is doing a lot of things and might benefit from being broken up."

Caveman Review:

L88-140: 🔵 nit: 50-line fn does 4 things. Extract validate/normalize/persist.

The extension uses Gemini CLI hooks (SessionStart and BeforeAgent) to inject the GEMINI.md rules and track your requested intensity level, ensuring the AI consistently applies the correct level of terseness across your session.