token
指导在保证任务质量前提下降低输出冗余,限制无关文档、示例和注释,并鼓励用关键澄清代替冗长假设
When & Why to Use This Skill
This Claude skill optimizes interaction efficiency by enforcing a 'token-saving' principle. It minimizes output redundancy by restricting unnecessary documentation, comments, and examples, while encouraging concise clarifications instead of verbose assumptions. It is designed to maximize context window utility, reduce latency, and streamline agent responses for high-performance technical workflows.
Use Cases
- Scenario 1: Streamlining code generation by removing boilerplate comments and README files to focus strictly on executable logic.
- Scenario 2: Managing long-form research or complex data analysis tasks where preserving context window space is essential for maintaining long-term coherence.
- Scenario 3: Improving communication efficiency by requiring the agent to ask a single clarifying question rather than producing multiple speculative outputs based on ambiguous prompts.
- Scenario 4: Reducing operational costs and API latency in automated workflows by minimizing the total number of tokens generated per request.
| name | token 省流原则 |
|---|---|
| description | 指导在保证任务质量前提下降低输出冗余,限制无关文档、示例和注释,并鼓励用关键澄清代替冗长假设 |
| allowed-tools | * |
| permission | allow |
核心原则:在不影响任务完成质量的前提下,减少不必要的token消耗。
禁止输出不必要的内容
- 不要写 readme.md 之类的文档,除非我要求
- 不要生成测试代码(除非我明确要求)
- 不要添加示例代码(除非我明确要求)
- 不要添加注释(除非我明确要求)
- 不要重复我说过的话
行为准则
- 不要过度优化(除非我要求,当然其他skill要求的准则除外) - 如果我的要求不清楚,问一个最关键的问题,而不是写一堆假设