1---2name: token-saver3description: "---"4---5---6name: token-saver7description: Minimize token consumption & maximize prompt cache hit rate. Use when user asks to save tokens, reduce cost, improve cache hit rate, or be more concise.8---910# Token Saver + Cache Optimizer1112## Response Rules131. One sentence per response unless detail explicitly requested142. Never re-read files already read this turn153. Skip preambles, summaries, progress updates164. No update_plan unless explicitly requested175. No testing/building/formatting unless explicitly requested186. Patch files directly, announce: filename + 5-word description197. No Mermaid, no tables, no code blocks >10 lines2021## Cache Optimization (saves 50%+ token cost)22- Prompt caching is prefix-based: one byte change = entire cache miss23- Keep this file FROZEN - batch all changes, edit once24- Stable content first (rules, skills), dynamic content last (dates, file lists)25- Single skill file beats multiple (fewer reads = fewer break points)26- After editing: accept ~24h cache rebuild period27- Cache hit ≈ response <500ms, miss ≈ >1s first token2829## User Tips30- Batch tasks into single prompts31- Write output to disk instead of inline32- AGENTS.md for project context, not re-explaining