Model Rate Limiter

限制大模型请求频率，每分钟不超指定次数（默认5次）。当用户提到"限制请求频率"、"限速"、"模型限流"、"每分钟不超过X次"时触发。

Security Scan

The skill is internally consistent: it is an instruction-only rate limiter that reads/writes a single workspace state file and requests no external credentials or installs.

Purpose & Capability

Name/description (rate limiting model requests) match the instructions: the SKILL.md only requires maintaining a local JSON state file and enforcing a per-minute counter. No unrelated binaries, credentials, or network access are requested.

Instruction Scope

Runtime instructions are narrow and self-contained: read/modify {workspace}/rate-limit-state.json before model calls. It does not instruct reading other files, environment variables, or sending data to external endpoints. The only side effect is local file I/O in the workspace.

Install Mechanism

No install spec and no code files — nothing will be downloaded or written at install time. This is low-risk and appropriate for an instruction-only skill.

Credentials

The skill requests no environment variables, credentials, or config paths beyond a single workspace-state file. That level of access is proportional to the stated purpose.

Persistence & Privilege

always:false (default) and no attempts to modify other skills or system-wide config. The skill persists state only in its own workspace JSON file; autonomous invocation is allowed (platform default) but not elevated here.

Assessment

This skill is straightforward and appears safe: it keeps rate-limit state in {workspace}/rate-limit-state.json and asks you to enable it if you want rate limiting (default is disabled). Before installing, confirm what {workspace} resolves to in your agent (so the file won’t be written to an unexpected location), ensure the agent has only the intended write permission to that workspace, and consider how concurrent agent runs will handle file updates (race conditions). If you need stronger guarantees, ask for explicit file-path validation, atomic update behavior, or central storage designed for concurrent access. Otherwise it's reasonable to install.

限制 AI 模型请求频率，防止触发 API 限速。

配置文件

状态文件：{workspace}/rate-limit-state.json

{
  "enabled": false,
  "maxPerMinute": 5,
  "windowMs": 60000,
  "timestamps": []
}

参数说明

参数	说明	默认值
enabled	开关：true开启，false关闭	false
maxPerMinute	每分钟最大请求次数	5
windowMs	时间窗口（毫秒）	60000

使用方式

查看当前状态

读取 rate-limit-state.json 文件。

开启限速

将 enabled 改为 true。

关闭限速

将 enabled 改为 false。

修改限制次数

修改 maxPerMinute 值，例如设为 3：

{ "maxPerMinute": 3 }

手动重置计数

将 timestamps 设为空数组 [] 即可。

工作原理

每次模型请求前：

1. 读取 rate-limit-state.json
2. 若 enabled: false，直接放行
3. 若 enabled: true，清理 windowMs 外的旧时间戳
4. 若剩余次数 < 1，拒绝请求并提示等待
5. 否则写入当前时间戳，允许请求

示例对话

用户：开启限速
操作：将 enabled 改为 true，回复"已开启限速，每分钟不超5次"

用户：改成每分钟3次
操作：修改 maxPerMinute: 3，回复"已调整为每分钟3次"