← 返回 Skills 市场
847
总下载
0
收藏
3
当前安装
1
版本数
在 OpenClaw 中安装
/install ollama-model-tuner
功能描述
Locally fine-tune Ollama models, prompts, and LoRAs using custom datasets and evaluation metrics without requiring cloud resources.
安全使用建议
This package is misleading/incomplete: it advertises fine‑tuning and LoRA support but only contains a small inference script. Before installing or running it: (1) ask the author for the missing files (prompts/system.md) and a real CLI or entrypoint, (2) verify/obtain clear install instructions for the 'ollama' Python client and any dependencies, (3) review any additional code the author provides to confirm actual training steps, (4) run the code in an isolated environment (container/VM) and with non‑sensitive data, and (5) prefer skills with a verifiable source/homepage or published install mechanisms. If you need true local fine‑tuning, use a vetted tool or official Ollama documentation rather than this incomplete skill.
功能分析
Type: OpenClaw Skill
Name: ollama-model-tuner
Version: 1.0.0
The skill bundle contains a Python script (`scripts/tune.py`) that is vulnerable to Local File Inclusion (LFI). The script directly uses `sys.argv[2]` as a file path for `dataset_file` without sanitization, allowing an attacker to read arbitrary files on the system by providing a malicious path. This critical vulnerability allows unauthorized data access, classifying the skill as suspicious.
能力评估
Purpose & Capability
The README/description claims fine‑tuning, LoRA tuning, and benchmarking, but the provided script (scripts/tune.py) only runs inference (ollama.chat) over up to 10 samples. Promised capabilities (training/LoRA application/evaluation metrics) are not implemented. SKILL.md references prompts/system.md and a CLI name (!ollama-model-tuner) that do not exist in the manifest.
Instruction Scope
Runtime instructions point users to run a CLI (!ollama-model-tuner) and reference files that aren't included. The SKILL.md does not explain how to install or provide the executable, so following the instructions will likely fail. The instructions do not request sensitive system files or extra credentials (no evidence of exfiltration), but their vagueness grants broad discretion and is inconsistent with the actual code.
Install Mechanism
There is no install spec (instruction-only), which is low risk, but the included Python script depends on the 'ollama' Python package and a running Ollama service; no dependency or install guidance is provided. This mismatch means the skill may not run as described and could lead users to run ad-hoc install commands from unknown sources.
Credentials
The skill requests no environment variables, credentials, or config paths. The only runtime dependency is the 'ollama' Python client which typically communicates with a local Ollama daemon; this is proportionate to the described local usage.
Persistence & Privilege
The skill is not always-on and does not request elevated or persistent system privileges. It does not modify other skills or system-wide settings in the provided files.
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 ollama-model-tuner - 安装完成后，直接呼叫该 Skill 的名称或使用
/ollama-model-tuner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
v1 launch: Local Ollama prompt/model tuning.
元数据
常见问题
Ollama Model Tuner 是什么？
Locally fine-tune Ollama models, prompts, and LoRAs using custom datasets and evaluation metrics without requiring cloud resources. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 847 次。
如何安装 Ollama Model Tuner？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install ollama-model-tuner」即可一键安装，无需额外配置。
Ollama Model Tuner 是免费的吗？
是的，Ollama Model Tuner 完全免费（开源免费），可自由下载、安装和使用。
Ollama Model Tuner 支持哪些平台？
Ollama Model Tuner 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 Ollama Model Tuner？
由 Goroni（@gblockchainnetwork）开发并维护，当前版本 v1.0.0。
推荐 Skills