Caveman mode¶
Optional response compression for agent chat after sync. Compresses style, not substance. Does not change sync or validation.
Inspired by JuliusBrussee/caveman — this toolkit ships a portable preferences/contract, not a full port. See Credits.
Default¶
OFF. Prefs live at {{SDD_ROOT}}/preferences.json after sync (caveman_mode, caveman_level).
Activate / deactivate¶
| Command | Effect |
|---|---|
caveman on |
Enable |
caveman off |
Disable |
caveman status |
Report on/off + level |
caveman lite | full | ultra |
Set level (turns ON if off) |
stop caveman / normal mode |
Same as off |
Levels and caps¶
- lite — tight full sentences
- full — default when ON; fragments OK
- ultra — maximum compression
Skills may cap intensity (planning often Lite; develop Full). NEVER compress for help-skills, commit, push, open-github-pr.
Auto-Clarity¶
Drop compression for security warnings, irreversible confirms, ambiguous multi-step order, or when the user asks to clarify — then resume.
When to use¶
Prefer ON for long review / orchestration. Prefer OFF for short Q&A (loading the contract has a cost).
Full guide: docs/guides/07-caveman-mode.md · contract: core/skills/_shared/caveman/CAVEMAN.md.
Next: Using skills · Credits · Home