← 返回 Skills 市场
RunPod pod management
作者
andrewharp
· GitHub ↗
· v1.0.1
1366
总下载
0
收藏
3
当前安装
2
版本数
在 OpenClaw 中安装
/install runpod
功能描述
Manage RunPod GPU cloud instances - create, start, stop, connect to pods via SSH and API. Use when working with RunPod infrastructure, GPU instances, or need SSH access to remote GPU machines. Handles pod lifecycle, SSH proxy connections, filesystem mounting, and API queries. Requires runpodctl (brew install runpod/runpodctl/runpodctl).
安全使用建议
What to check before installing: 1) This skill expects runpodctl, ssh, sshfs (FUSE) and a RunPod API key — the registry metadata doesn't list those binaries, so make sure you have them and trust runpodctl's source. 2) The mount script will read your SSH private key (from ~/.runpod/ssh/ or a path you set via RUNPOD_SSH_KEY) and write a known_hosts file under ~/.runpod/ssh/; mounting gives the remote root filesystem access to your local mountpoint — only mount pods you trust. 3) The script parses runpodctl output using grep -P (Perl regex); this may be non-portable on some systems (macOS grep may not support -P). 4) Host key handling uses StrictHostKeyChecking=accept-new (it will accept new host keys automatically and store them in the skill’s known_hosts file) — be aware of the security tradeoff. If you’re comfortable with these points and trust RunPod/runpodctl, the skill appears coherent. If unsure, inspect runpodctl and run the mount script in a sandboxed environment first.
功能分析
Type: OpenClaw Skill
Name: runpod
Version: 1.0.1
The skill is classified as suspicious due to its use of high-risk capabilities, specifically the `scripts/mount_pod.sh` script. This script utilizes `sshfs` to mount the remote RunPod instance's root filesystem (`root@$IP:/`) locally, which grants extensive file system access. It also handles SSH keys (`IdentityFile`) and uses `StrictHostKeyChecking=accept-new` with a dedicated `known_hosts` file, which, while plausibly needed for managing dynamic cloud environments, represents a significant security capability. There is no clear evidence of intentional malicious behavior, data exfiltration, or prompt injection attempts against the agent; the risky behaviors are aligned with the stated purpose of managing and interacting with RunPod instances.
能力评估
Purpose & Capability
The skill is focused on RunPod pod lifecycle and SSH/SSHFS mounting. The SKILL.md and the mount script implement that functionality (calling runpodctl, using SSH keys, sshfs). However, the registry metadata declares no required binaries/env vars even though the instructions require runpodctl, ssh, and sshfs/fusermount; this appears to be an oversight rather than malicious mismatch.
Instruction Scope
Instructions are limited to RunPod operations: installing runpodctl, configuring an API key via runpodctl, managing SSH keys, and mounting pods via the included script. The script reads/writes SSH key and known_hosts files under ~/.runpod or ~/.ssh (expected for SSH/sshfs). It does parse runpodctl output (grep -oP) which is brittle and may fail on systems where grep lacks -P support; otherwise there is no scope creep (no unrelated system-wide reads or exfil endpoints).
Install Mechanism
This is an instruction-only skill with a small helper script; there is no install spec or remote download, so nothing arbitrary is written to disk by an installer. The included shell script is the only code and its behavior is visible.
Credentials
The skill declares no required environment variables or credentials, but the SKILL.md asks users to run 'runpodctl config --apiKey' (so an API key is required for RunPod use) and documents optional env vars RUNPOD_SSH_KEY and RUNPOD_KNOWN_HOSTS used by the script. The lack of declared required creds in metadata is a minor inconsistency but not an indication of hidden credential access.
Persistence & Privilege
The skill does not request permanent always-on inclusion and does not modify other skills or system-wide agent settings. The script will create directories under your home and write a per-skill known_hosts file (expected for SSH/sshfs usage).
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 runpod - 安装完成后，直接呼叫该 Skill 的名称或使用
/runpod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
v1.0.1
Updated scripts/mount_pod.sh to address warning about stricthostchecking. SSH keys for known pods will now be stored in ~/.runpod/ssh/known_hosts on first connect and verified thereafter.
v1.0.0
Initial release – RunPod GPU cloud instance management skill.
- Manage pod lifecycle: create, start, stop, and query RunPod GPU instances.
- Easily connect to pods via SSH using runpodctl.
- Supports SSHFS filesystem mounting with helper script.
- Includes quick reference for common runpodctl commands.
- Provides instructions for SSH key setup and configuration.
- Details on accessing web services running inside pods via RunPod proxy URLs.
元数据
常见问题
RunPod pod management 是什么？
Manage RunPod GPU cloud instances - create, start, stop, connect to pods via SSH and API. Use when working with RunPod infrastructure, GPU instances, or need SSH access to remote GPU machines. Handles pod lifecycle, SSH proxy connections, filesystem mounting, and API queries. Requires runpodctl (brew install runpod/runpodctl/runpodctl). 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 1366 次。
如何安装 RunPod pod management？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install runpod」即可一键安装，无需额外配置。
RunPod pod management 是免费的吗？
是的，RunPod pod management 完全免费（开源免费），可自由下载、安装和使用。
RunPod pod management 支持哪些平台？
RunPod pod management 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 RunPod pod management？
由 andrewharp（@andrewharp）开发并维护，当前版本 v1.0.1。
推荐 Skills