soleri-terselisted

# 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 blo