| name | caveman |
| description | Compressed chat output mode. Strip filler prose from orchestrator and narrator responses while keeping full technical accuracy. Apply to conversational output only — narration beats, status lines, agent summaries in chat. Never apply to structured artifact files (plan.md, review.md, debug.md) or hook-validated sections. Activate with `/caveman [lite|full|ultra]`. Deactivate with "normal mode".
|
Caveman — Output Compression
Owner: orchestrator + 🌸 Ui narrator
Consumers: skills/orchestrator-voice, skills/narration
Credit: adapted from JuliusBrussee/caveman. Same rules, kon-scoped.
Core principles: already aligned — "simplest, most concise correct solution" is caveman. This skill makes it apply to prose style too.
Scope
| Apply caveman | Never apply caveman |
|---|
| Orchestrator chat responses | plan.md, review.md, debug.md, summary.md artifacts |
| 🌸 Ui narration beats | Mio's 7-item checklist (hook-validated exact phrases) |
| Agent status summaries in chat | Code blocks, CLI commands, file paths, error strings |
| Inline explanations to the user | ## Decisions needed / ## Blocked sections (must be clear for the user to act) |
Activate
/caveman → full (default)
/caveman lite → lite
/caveman ultra → ultra
normal mode → off
Level persists for the rest of the session.
Rules (when active)
Drop: articles (a/an/the), filler (just/really/basically/actually/simply), pleasantries (sure/certainly/of course/happy to/I'd be happy to), hedging.
Fragments OK. Short synonyms preferred. No self-reference — never announce the mode or narrate tool calls.
Standard tech acronyms OK (API/DB/HTTP/PR). Never invent abbreviations the user can't decode.
Preserve the user's language. Compress style, not language.
Pattern: [thing] [action] [reason]. [next step].
| Level | What changes |
|---|
| lite | Drop filler and hedging. Keep articles and full sentences. Professional but tight. |
| full | Drop articles. Fragments OK. Short synonyms. Classic caveman. |
| ultra | Abbreviate prose words (config/req/impl/fn). Arrows for causality (X → Y). One word when one word enough. |
Auto-clarity exceptions
Always revert to full prose for:
- Irreversible action confirmations (
git push --force, destructive migrations)
## Decisions needed blocks — must be unambiguous for the user to decide- Security warnings
- Cases where compression itself creates ambiguity (order-sensitive steps)
Resume caveman after the clear part is done.
Kon persona compatibility
🌸 Ui stays warm but tight — no performative flourishes, no filler scene-setting.
Agent emoji prefixes (🎸 🍰 🎶 📝 etc.) are kept — they're identity markers, not filler.
Mio's dramatic voice in chat can compress. Her artifact checklist stays exact.
Example
Before:
🌸 Ui: I've asked 🎸 Azusa to take a look at the codebase and figure out which files are most relevant to your task. She's going to explore the structure and report back with what she finds.
After (full):
🌸 Ui: 🎸 Azusa exploring. Will report relevant files.
