with one click
token-saving
节省 token 消耗的最佳实践。USE FOR: 优化 prompt 长度、减少重复输出、控制上下文大小、提高成本效益。
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Menu
节省 token 消耗的最佳实践。USE FOR: 优化 prompt 长度、减少重复输出、控制上下文大小、提高成本效益。
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
qa-engineer 的默认 skill。USE FOR: 测试用例设计、执行验证、测试报告输出、缺陷分级。
unite-test 的默认 skill。USE FOR: 测试用例设计、执行验证、测试报告输出、缺陷分级。
developer 的默认 skill。USE FOR: 按设计实现代码、补齐测试、控制技术债、保证可运行性。
Use for Android middleware, AOSP, Android Automotive, Binder, AIDL, HIDL, HAL, SystemServer, CarService, Service, APK and framework-native boundary architecture from requirements and design documents.
Generate, revise, 补全 or重写 car middleware, cockpit middleware, IVI middleware architecture from PRD,需求文档,requirements/requirements_spec.md or feature specs. Use for Linux/QNX/Android cross-OS 架构设计, 中间件分层, 服务拓扑, 通信矩阵, 资源预算, 诊断, 安全 and design/architecture.md.
Use for C++ middleware, native service, performance critical module, real-time path, memory ownership, threading, IPC bridge, ABI boundary and reliability architecture from automotive or embedded requirements and design docs.
| name | token-saving |
| description | 节省 token 消耗的最佳实践。USE FOR: 优化 prompt 长度、减少重复输出、控制上下文大小、提高成本效益。 |
| globs | ["**/*.py","**/*.md","**/*.yaml"] |
本 Skill 帮助在保证输出质量的前提下最小化 token 消耗。
# ✅ 好:关键逻辑一行注释
def process(data):
# 跳过空值避免下游 NPE
return [x for x in data if x]
# ❌ 差:冗长解释
def process(data):
# This function takes a list of data items and processes them
# by filtering out any items that are None or empty. This is
# important because downstream functions expect non-null values
# and will throw NullPointerException if they receive null.
return [x for x in data if x]
req → requirements, impl → implementation# ✅ 好
实现 bubble_sort(arr),返回升序数组。
# ❌ 差
请帮我实现一个冒泡排序算法。这个算法应该接收一个数组作为输入,
然后通过比较相邻元素并交换位置的方式,最终返回一个从小到大排序好的数组。
## Context (if needed) 明确标记可选上下文# 明确限制输出范围
输出要求:
- 仅返回修改的函数,不输出未变更代码
- 每个文件 diff 不超过 50 行
- 跳过样板代码说明
// ... existing code ... 标记省略部分# ✅ 批量操作
files = ["a.py", "b.py", "c.py"]
for f in files:
edit(f, changes)
# ❌ 单独操作(每次调用都消耗 token)
edit("a.py", change1)
edit("b.py", change2)
edit("c.py", change3)
# 发现明显问题立即报告,不继续深入
if critical_issue:
return "VERDICT: FAIL - 缺少必要的错误处理"
参考 budget.yaml 配置:
当前用量:检查 /api/runtime 返回的 budget 字段
在完成任务前确认: