← 返回 Skills 市场
781
总下载
0
收藏
4
当前安装
1
版本数
在 OpenClaw 中安装
/install offline-llama
功能描述
Manage local Ollama models autonomously with health monitoring, automatic fallback, self-healing, and offline operation without internet dependency.
安全使用建议
This skill could do exactly what it claims, but it currently leaves important details out and grants broad, ambiguous authority. Before installing or enabling: 1) Verify the skill's source/author (unknown here). 2) Require the author to specify exact commands, required binaries (e.g., ollama CLI, systemctl), config paths, and any network endpoints used for model downloads. 3) Restrict the agent's runtime privileges (run in a sandbox or VM) so it cannot restart system services or modify files without explicit user confirmation. 4) Disable autonomous invocation or limit it to read-only health checks until you trust the implementation. 5) Prefer skills that declare required binaries/paths and provide an install script or code you can review. If you want to proceed, test in an isolated environment and monitor logs/changes closely.
功能分析
Type: OpenClaw Skill
Name: offline-llama
Version: 1.0.0
The skill, described in SKILL.md, outlines autonomous system-level actions such as restarting services (Ollama), clearing system resources (cache, temporary files), and reinstalling or downloading models from remote sources. While these capabilities are aligned with the stated purpose of managing local Ollama models, they grant significant control to the AI agent over the host system. The automatic triggering of these high-risk behaviors (service management, file system modification, network-based software installation) without explicit malicious instructions makes the skill suspicious, as these operations could be exploited if the underlying implementation contains vulnerabilities, or if the remote model sources are compromised.
能力评估
Purpose & Capability
The skill claims to manage Ollama models (health checks, restarts, reinstallations, cache clearing). Performing these tasks normally requires specific binaries/CLIs (e.g., ollama CLI, systemctl or init scripts), filesystem paths, and potentially network access. The skill declares no required binaries, config paths, or credentials, which is inconsistent with the stated capabilities.
Instruction Scope
SKILL.md gives broad runtime instructions (continuous monitoring, restarting services, clearing caches, reinstalling models, log analysis) but does not specify exact commands, files, or limits. The instructions are open-ended about which logs/files to read and allow autonomous decisions (e.g., when to reinstall), granting the agent wide discretion to access system state and perform potentially destructive actions.
Install Mechanism
No install spec and no code files are present (instruction-only). That minimizes risk from arbitrary downloads or written artifacts, but it also means the SKILL.md is the sole runtime authority — increasing importance of clear, constrained instructions which are currently lacking.
Credentials
The skill requests no environment variables or credentials, which superficially limits exfiltration risk. However, the described behaviors (reinstalling models, switching to remote models when internet is present) imply network access and possibly access to model registries; the lack of declared requirements or credential needs is an omission and reduces transparency about what privileges the agent will need.
Persistence & Privilege
The skill is not forced-always and allows normal autonomous invocation. Autonomous invocation combined with system-management actions increases blast radius if misused, but 'always: false' and default invocation settings are reasonable. There's no evidence it attempts to modify other skills or system-wide configs from the provided text.
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 offline-llama - 安装完成后，直接呼叫该 Skill 的名称或使用
/offline-llama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: Enables autonomous, offline-first management and health monitoring of local Ollama models.
- Monitors local model health with automatic fallback and dynamic model switching.
- Includes self-healing: restarts Ollama, clears resources, reinstalls models if needed.
- Maintains continuous operation and degraded mode when all models are unavailable.
- Detects internet connectivity to switch between local and remote models when necessary.
- Provides commands for manual control (status checks, switching, restarting, clearing cache).
- Fully local operation—no external dependencies, preserving privacy.
元数据
常见问题
Offline Llama 是什么？
Manage local Ollama models autonomously with health monitoring, automatic fallback, self-healing, and offline operation without internet dependency. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 781 次。
如何安装 Offline Llama？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install offline-llama」即可一键安装，无需额外配置。
Offline Llama 是免费的吗？
是的，Offline Llama 完全免费（开源免费），可自由下载、安装和使用。
Offline Llama 支持哪些平台？
Offline Llama 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 Offline Llama？
由 and-ray-m（@and-ray-m）开发并维护，当前版本 v1.0.0。
推荐 Skills