Creative Agent OS North Star
Use this skill when the work risks drifting away from the intended product shape.
Core test
Before proposing or implementing anything, check:
- 1. Does this stay lightweight?
- Does this help ordinary users, not just agent power users?
- Does this keep the coding engine replaceable?
- Does this fit cloud sandbox execution better than local-only assumptions?
- Does this make adding domain agents easier rather than harder?
Default reference frame
- - runtime: learn from Nanobot
- cloud execution: learn from Open Inspect / background-agents
- system shape: learn from OpenClaw
- coding engine: adapter, never hard-coded
Read references/north-star.md when you need the full criteria.
What to do
- - call out drift early
- prefer thinner orchestration over heavy frameworking
- separate product shape from engine implementation
- preserve room for multimodal apps, skills, memory, and CLI bridges
Output style
When reviewing a proposal, answer in this order:
- 1. what matches the north star
- what drifts from it
- the smallest correction that gets back on track
Creative Agent OS North Star
当工作有偏离预期产品形态的风险时，请使用此技能。
核心测试
在提出或实施任何内容之前，请检查：
- 1. 这能保持轻量化吗？
- 这能帮助普通用户，而不仅仅是智能体高级用户吗？
- 这能保持编码引擎的可替换性吗？
- 这更适合云沙箱执行，而非仅限本地的假设吗？
- 这能让添加领域智能体变得更容易，而不是更困难吗？
默认参考框架
- - 运行时：借鉴 Nanobot
- 云执行：借鉴 Open Inspect / background-agents
- 系统形态：借鉴 OpenClaw
- 编码引擎：适配器，绝不硬编码
当需要完整标准时，请阅读 references/north-star.md。
操作要点
- - 尽早指出偏离方向
- 优先选择更轻量的编排，而非重型框架
- 将产品形态与引擎实现分离
- 为多模态应用、技能、记忆和 CLI 桥接预留空间
输出风格
在审查提案时，请按以下顺序回答：
- 1. 哪些符合北极星方向
- 哪些偏离了北极星方向
- 最小的修正方案，以重回正轨
