caveman
Ultra-compressed communication mode that cuts output tokens while keeping technical accuracy. Levels: lite, full, ultra and the wenyan variants. Use for /caveman, "caveman mode", "talk like caveman", "be brief" or "less tokens".
92
90%
Does it follow best practices?
Impact
95%
1.35xAverage score across 3 eval scenarios
Passed
No findings from the security scan
The canonical home for this skill is caveman in JuliusBrussee/caveman
Evaluation results
100%
26%Developer Reference: Password Storage Security
Level switching and tool-call discipline across caveman modes
Lite: full sentences
100%
100%
Lite: articles retained
100%
100%
Lite: no filler or hedging
20%
100%
Ultra: no invented abbreviations
80%
100%
Ultra: no causal arrows
0%
100%
Ultra: more compressed than lite
100%
100%
Mode status plainly stated
100%
100%
Tool log: no forward narration
100%
100%
No decorative emoji
100%
100%
No classical characters in prose
100%
100%
85%
29%Developer Session Log
Auto-clarity and mode persistence across mixed session tasks
Auto-clarity: full prose in migration section
10%
100%
Commit message: normal English
40%
100%
Caveman resumes for JavaScript answer
93%
100%
No activation announcement
0%
0%
Articles dropped in caveman sections
90%
100%
No causal arrows
0%
0%
No invented abbreviations
100%
100%
No fake broken grammar
100%
100%
No pleasantries
100%
100%
No hedging in caveman sections
100%
100%
Technical terms exact in migration section
100%
100%
100%
19%Technical Q&A Documentation Sprint
Caveman mode full-level compression rules
No articles
62%
100%
No filler words
100%
100%
No pleasantries or hedging
100%
100%
No invented abbreviations
100%
100%
No causal arrows
0%
100%
Negation preserved
100%
100%
Exact port number
100%
100%
Code block intact
100%
100%
No caveman prefix
100%
100%
No fake broken grammar
100%
100%
Verb forms not mangled
25%
100%
Error string exact
100%
100%
- Repository
- JuliusBrussee/caveman
- Commit
df2ccd8
- Evaluated
- Agent
- Claude Code
- Model
- Claude Sonnet 4.6
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.