Skip to content

AlexChen31337/terse

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 

Repository files navigation

terse

🪨 why use many token when few token do trick

Compressed output mode for AI agents. Cuts ~65–75% of output tokens by stripping filler words, pleasantries, articles, and hedging — while keeping code, technical terms, and error messages verbatim.

Based on caveman.

Token Savings

Task Normal Terse Saved
React re-render bug 1180 159 87%
PostgreSQL pool setup 2347 380 84%
Git rebase conflict 891 374 58%

March 2026 paper: brevity constraints improved accuracy by 26pp.

Levels

  • lite — Drop filler, hedging. Full sentences.
  • full — Omit articles, use fragments, bare imperatives.
  • ultra — Max compress. Labels only. No sentences.

Quick Start

# Install as OpenClaw skill
clawhub install terse

# Or manual: copy to ~/.openclaw/workspace/skills/terse/

CLI helper

uv run python scripts/caveman_prompt.py --level full "your task here"

In sub-agent prompts

CAVEMAN MODE: Omit articles, filler, pleasantries. Use fragments.
Steps as bare imperatives. Keep code/errors verbatim. No apologies.
No "I". Just signal.

[your task]

Model Pairing

Level Best model Why
Lite Any Minimal overhead
Full Sonnet 4.6 Follows compression well
Ultra Haiku 4.5 Cheap + short = efficient

License

MIT

About

🪨 Compressed output mode for AI agents. why use many token when few token do trick. ~65-75% token savings.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages