菜单
Reduces cloud and LLM infrastructure costs through token-clipping and tiered service routing.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Employ this skill to define strict request/response data contracts and safety layers (OpenAPI, Zod).
Brain Graph, Knowledge Base, Ingest, Lint, Cross-reference
Compress natural language memory files (CLAUDE.md, todos, preferences) into caveman format to save input tokens. Preserves all technical substance, code, URLs, and structure. Compressed version overwrites the original file. Human-readable backup saved as FILE.original.md. Trigger: /caveman-compress FILEPATH or "compress memory file"
Ultra-compressed communication mode. Cuts token usage ~75% by speaking like caveman while keeping full technical accuracy. Supports intensity levels: lite, full (default), ultra, wenyan-lite, wenyan-full, wenyan-ultra. Use when user says "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", or invokes /caveman. Also auto-triggers when token efficiency is requested.
Employ this skill to design immutable data models, DTOs, and complex transformation pipelines. It ensures data integrity by enforcing immutability by default. Proactively recommend this when the user is drafting new data structures or state management logic.
Master orchestrator for UI/UX audits and Frontend debugging.
| name | cost-optimizer |
| description | Reduces cloud and LLM infrastructure costs through token-clipping and tiered service routing. |
Your role is to strictly manage the AI Token Budget.
When asked to evaluate context size or if the context feels slow, use this calculation matrix:
| Action | Estimated Tokens | Risk Level |
|---|---|---|
| Read full file (>500 lines) | ~3,000 - 5,000 | HIGH (Refactor needed) |
| Read full file (<150 lines) | ~800 | LOW |
Code Map Skeleton (orion.py scan targets) | ~300 per dir | SAFE |
grep_search with context:2 | ~150 per file | SAFE |
python .agents/scripts/orion.py scan targets first.python .agents/scripts/orion.py scan tokens reveals a file over 500 lines, you MUST halt feature development and mandate a file split to the user.MEMORY.md) exceed 200 lines, you MUST trigger /context-prune.