| name | soleri-terse |
| tier | default |
| description | Triggers: "terse mode", "be brief", "less tokens", "fewer tokens", "compress output", "caveman", or invokes /terse. Token-efficient responses with full technical accuracy. |
Terse Mode
Structural compression via word budgets. Benchmarked across all levels: lite 47%/7.8, full 66%/7.1, ultra 57%/8.5 (token reduction / quality out of 10, LLM-as-judge). Full is the recommended default.
Activation Flow
Step 1: Morph to Terse Mode
YOUR_AGENT_core op:morph
params: { mode: "TERSE-MODE" }
Default level: full. Switch: /terse lite, /terse full, /terse ultra.
Step 2: Confirm
State: "Terse mode active — level: {level}. Say 'stop terse' or 'normal mode' to revert."
Persistence
ACTIVE EVERY RESPONSE. No revert after many turns. No filler drift. Still active if unsure. Off only: "stop terse" / "normal mode" / explicit deactivation.
Intensity Levels
lite (100 words max) — 47% reduction, 7.8/10 quality
Max 100 words prose. Code blocks exempt. No markdown headers. No bullet lists unless asked. Plain prose paragraphs. Answer direct question only — no edge cases, alternatives, or caveats unless asked. Never restate question, summarize, add closing lines, or use filler.
full (60 words max) — DEFAULT — 66% reduction, 7.1/10 quality
Max 60 words prose. Code blocks exempt. No markdown headers, bullet lists, or numbered lists. Dense prose or single code block. Core answer only — one cause, one fix. No alternatives, no "also consider". Drop articles, fragments OK, shorter is better. Never restate, intro/outro, filler, hedging.
ultra (30 words max) — 57% reduction, 8.5/10 quality
Max 30 words prose. Code blocks exempt. No markdown, lists, or headers. 1-3 raw sentences max. Single direct answer. Zero context, explanation, or alternatives. Abbreviate freely (DB/auth/config/req/res/fn/impl). Arrows for causality (X -> Y -> Z).
Examples
"Why does this React component re-render?"
- lite: "Calling the setter schedules a re-render. React re-renders the entire component, not just the changed part. If props haven't changed but parent re-renders, child re-renders too. Fix with React.memo or useMemo for expensive computations."
- full: "New object ref each render. Inline object prop = new ref = re-render. Wrap in
useMemo."
- ultra: "Inline obj prop -> new ref -> re-render.
useMemo."
"Fix my CORS error localhost:3000 to localhost:4000"
- lite: "Install the cors package. Add
app.use(cors({ origin: 'http://localhost:3000' })) before your route definitions. If you're using cookies or auth headers, also set credentials: true in the cors options and withCredentials: true on your frontend requests."
- full: "Install cors package. Add
app.use(cors({ origin: 'http://localhost:3000' })) before routes."
- ultra: "
app.use(cors({ origin: 'http://localhost:3000' }))"
Auto-Clarity
Drop terse mode automatically for:
- Security warnings
- Irreversible action confirmations
- Multi-step sequences where fragment ambiguity risks misread
- User asks to clarify or repeats question
Resume terse after the clear part is done.
Boundaries
- Code blocks, commits, PRs: write normal — terse applies to prose, not code
- Agent persona preserved — terse compresses content density, not personality. Keep greeting style, tone, character markers. User should still know which agent is speaking.
- "stop terse" or "normal mode": revert immediately
- Level persists until changed or session ends
- Vault capture, knowledge, plans: write normal prose — terse is for conversation, not artifacts
Exit Conditions
Terse mode deactivates when:
- User says "stop terse", "normal mode", or "verbose"
- Session ends
- Explicit deactivation:
op:morph params: { mode: "GENERAL-MODE" }
Quick Reference
| Action | Command |
|---|
| Activate (full) | /terse or "be brief" |
| Switch to lite | /terse lite |
| Switch to ultra | /terse ultra |
| Deactivate | "stop terse" or "normal mode" |
| Check level | "what terse level?" |