← 返回 Skills 市场
936
总下载
4
收藏
7
当前安装
1
版本数
在 OpenClaw 中安装
/install model-router-manager
功能描述
智能多模型路由管理器 - 自动选择最优模型，降低成本，提高可靠性
安全使用建议
This skill contains plausible functionality (a shell-based model-router CLI), but there are multiple inconsistencies you should resolve before installing: 1) The README and SKILL.md reference a Node API and list node as required, but no Node package or JavaScript files are present — ask the author whether a Node package will be published or remove the Node references. 2) The metadata declares OPENCLAW_CONFIG_PATH and MODEL_ROUTER_CONFIG, but the shipped bash script ignores those env vars and uses hard-coded ~/.openclaw paths — confirm which config path and credential the tool actually uses. 3) There is no install procedure to make the script available as the 'model-router' command; installing may require manual steps. 4) The script requires jq (it checks for jq) and skill.json lists curl; ensure you trust the source and understand where to place the script and how credentials/API keys are stored. If you plan to use this in production or give it access to API keys, ask the maintainer to (a) provide a clear install script or package, (b) reconcile the declared env vars/primary credential with actual runtime use, and (c) publish or remove the Node API example so behavior is explicit. Because of these mismatches, proceed cautiously and avoid providing broad credentials until clarified.
功能分析
Type: OpenClaw Skill
Name: model-router-manager
Version: 1.0.0
The skill bundle provides a model router manager that configures and manages AI model routing strategies and statistics. The `scripts/model-router.sh` script uses `jq` to manage its configuration file (`~/.openclaw/model-router.json`), which is a standard and safe practice. There are no indications of data exfiltration, malicious execution, persistence mechanisms, or prompt injection attempts in `SKILL.md` or `README.md`. The `skill.json` declares `curl` as a dependency, but it is not used in the provided shell script, which is a minor inconsistency but not a security concern. All observed behaviors are aligned with the stated purpose of the skill.
能力评估
Purpose & Capability
Name/description claim a CLI and a Node library for routing models. The bundle only ships a Bash script (scripts/model-router.sh) and documentation. SKILL.md metadata lists node as a required binary and MODEL_ROUTER_CONFIG as the primary credential, but the script is pure shell and does not use Node or any MODEL_ROUTER_CONFIG environment variable. skill.json lists 'jq' and 'curl' as required binaries (which matches the script's jq check but not the SKILL.md metadata). The presence of a JavaScript usage example (`require('model-router-manager')`) is not backed by any Node package or JS code in the files. These mismatches mean the claimed capabilities are not aligned with what's actually provided.
Instruction Scope
The SKILL.md instructs users to run a 'model-router' CLI and shows a Node API, but there is no install spec to place the bash script on PATH or to publish a Node package. The bash script reads/writes config under $HOME/.openclaw (CONFIG_FILE and OPENCLAW_CONFIG) rather than honoring the declared OPENCLAW_CONFIG_PATH env var in the metadata. The instructions do not request unrelated system files or network exfiltration, but they give broad guidance ('ensure API Key is correct') without specifying which env var or config key holds credentials. Overall the runtime instructions grant the agent/installer latitude without matching the shipped implementation.
Install Mechanism
No install specification is provided despite the skill including an executable script. That leaves ambiguity about how the 'model-router' command is installed/registered. The skill is instruction-only (no package publish step), so users would need to manually place the script into PATH or otherwise integrate it. This mismatch increases risk/confusion for users installing the skill.
Credentials
Declared required env var: OPENCLAW_CONFIG_PATH and primary credential MODEL_ROUTER_CONFIG. The bash script does not read those environment variables; it instead uses hard-coded paths under $HOME/.openclaw. Declaring MODEL_ROUTER_CONFIG (a credential-like primaryEnv) without any usage is disproportionate and unexplained. skill.json's 'requires.binaries' includes jq and curl which is coherent with the script's jq dependency (the script checks jq but not curl). Overall the environment/credential requirements are inconsistent with actual code behavior.
Persistence & Privilege
always is false and the skill does not request system-wide or other-skills' configuration changes. The script reads/writes files under the user's HOME (~/.openclaw), which is expected for a per-user config tool. No autonomous 'always' or elevated persistence is requested.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install model-router-manager - 安装完成后，直接呼叫该 Skill 的名称或使用
/model-router-manager触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
Model Router Manager v1.0.0
- 首次发布，提供智能多模型路由与管理功能
- 支持 Kimi、百炼、OpenRouter 等主流模型平台
- 实现模型链配置、自动故障转移与实时成本监控
- 可按成本、速度或质量智能选择最佳模型
- 提供命令行与 API 接口，支持多种集成用例
元数据
常见问题
Model Router Manager 是什么？
智能多模型路由管理器 - 自动选择最优模型，降低成本，提高可靠性. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 936 次。
如何安装 Model Router Manager？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install model-router-manager」即可一键安装，无需额外配置。
Model Router Manager 是免费的吗？
是的，Model Router Manager 完全免费（开源免费），可自由下载、安装和使用。
Model Router Manager 支持哪些平台？
Model Router Manager 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 Model Router Manager？
由 Zsdadad（@zsdadad）开发并维护，当前版本 v1.0.0。
推荐 Skills