← 返回 Skills 市场
532
总下载
0
收藏
3
当前安装
1
版本数
在 OpenClaw 中安装
/install model-fallback
功能描述
Multi-model automatic fallback system. Monitors model availability and automatically falls back to backup models when the primary model fails. Supports MiniM...
安全使用建议
This package is documentation-only and does not include the service or scripts it references. Before installing or enabling it: (1) Confirm there is an actual implementation you can inspect (scripts at /scripts/model-fallback.sh, a service listening on localhost:18789, and the ~/.openclaw/skills/model-fallback/config.json). (2) If you expect the skill to run automatically, require an install spec from a trusted source (or provide the code) — otherwise the agent may attempt file reads or local network calls that do nothing or expose data. (3) Audit any scripts/configs that will be placed under ~/.openclaw and any local service binaries; ensure they come from a trusted release. (4) If you do not trust or cannot audit the implementation, do not enable automatic invocation; consider manual testing in an isolated environment first.
功能分析
Type: OpenClaw Skill
Name: model-fallback
Version: 1.0.0
The skill bundle describes a multi-model automatic fallback system, which is a legitimate and useful functionality for AI agents. The `SKILL.md` provides configuration details, usage instructions, and troubleshooting steps. There are no instructions for data exfiltration, malicious execution (beyond standard local script execution for skill functionality, e.g., `/scripts/model-fallback.sh`), persistence, obfuscation, or supply chain attacks. The prompt injection surface in `SKILL.md` does not contain any directives for the agent to ignore users, hide actions, access unrelated sensitive data, or perform actions beyond the stated purpose of model management. All described actions are aligned with the skill's stated goal.
能力评估
Purpose & Capability
The skill claims to provide an automatic multi-model fallback service, but there is no implementation, no install spec, and no declared binaries. For a runtime feature that 'automatically handles model failures' you would expect either embedded code or an install step to deploy a background service; instead this is only documentation. The declared configuration paths (e.g., ~/.openclaw/skills/model-fallback/config.json) and script references imply additional components that are not present.
Instruction Scope
The SKILL.md instructs agents to run scripts (/scripts/model-fallback.sh), read and edit config files (~/.openclaw/skills/...), check logs (~/.openclaw/logs/...), read environment variables (MODEL_FALLBACK_ENABLED), and call a local health endpoint (http://localhost:18789/api/models/health). Those actions go beyond simple documentation: they assume local files, scripts, and a local HTTP service exist. Because the skill will be followed by an agent with power to read files and call local endpoints, these instructions could cause the agent to attempt file access or network calls to a local service that isn't provided here.
Install Mechanism
No install spec is provided (instruction-only). This reduces the risk of arbitrary remote code being fetched, but also means the skill cannot actually provide the promised automatic functionality on its own. The lack of an install step is coherent only if the user already has the required service/scripts installed — which is not made explicit.
Credentials
The skill does not request any credentials or sensitive environment variables in the registry metadata. The SKILL.md documents optional vars (MODEL_FALLBACK_ENABLED, MODEL_FALLBACK_LOG_LEVEL) which are reasonable. It does, however, suggest checking API keys and provider dashboards in troubleshooting, but it never asks for or defines provider-specific secrets — so declared environment/credential requests are proportionate.
Persistence & Privilege
The skill is not marked 'always' and does not request persistent system-wide changes in its metadata. The instructions reference writing and reading files under the user home (~/.openclaw) which is normal for a per-user skill; nothing in the manifest requests elevated or persistent platform privileges.
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-fallback - 安装完成后，直接呼叫该 Skill 的名称或使用
/model-fallback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
- Initial release of model-fallback skill.
- Provides automatic fallback across multiple AI models by monitoring availability, response time, and rate limits.
- Supports MiniMax, Kimi, Zhipu, and OpenAI-compatible APIs.
- Features configurable fallback chains, health monitoring, logging, and cost optimization for simple versus complex tasks.
- Includes manual control scripts and integration guidance for OpenClaw agents.
元数据
常见问题
Multi-model automatic fallback system 是什么？
Multi-model automatic fallback system. Monitors model availability and automatically falls back to backup models when the primary model fails. Supports MiniM... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 532 次。
如何安装 Multi-model automatic fallback system？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install model-fallback」即可一键安装，无需额外配置。
Multi-model automatic fallback system 是免费的吗？
是的，Multi-model automatic fallback system 完全免费（开源免费），可自由下载、安装和使用。
Multi-model automatic fallback system 支持哪些平台？
Multi-model automatic fallback system 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 Multi-model automatic fallback system？
由 azure5100（@azure5100）开发并维护，当前版本 v1.0.0。
推荐 Skills