← Back to Skills Marketplace
132
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install runninghub-skill
Description
RunningHub AI 智能调用。Use when user wants to generate images, videos, or audio content.
Usage Guidance
This skill is a wrapper around a separate project (RHMCP) and asks you to clone and run that external code. That means: 1) Inspect the RHMCP repository and the included executor.mjs before running npm install or node — running npm install/build executes arbitrary third-party code. 2) Be cautious where you set RHMCP_CONFIG and create service.json/.env — the executor reads files from those directories (apps.json, service.json) and could process any local config placed there. 3) The only declared credential is RUNNINGHUB_API_KEY (expected), but the code also uses RHMCP_CONFIG (not declared); avoid pointing it at directories that contain other secrets. 4) Prefer testing in an isolated environment (container or VM), and avoid running these install steps as root. 5) If you want minimal risk, request a version of the skill that calls RunningHub APIs directly (no external repo clone) or provide an official install package/release rather than free-form git clone + npm install.
Capability Analysis
Type: OpenClaw Skill
Name: runninghub-skill
Version: 1.0.0
The skill bundle is a legitimate integration for the RunningHub AI platform, providing an AI agent with the necessary logic and metadata to generate images, videos, and audio content. The SKILL.md file contains clear instructions for the agent on mapping user requests to specific tool calls and managing multi-step workflows, while the executor.mjs script serves as a helper for configuration management and decision-making. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found, and all components are consistent with the stated purpose of the skill.
Capability Assessment
Purpose & Capability
Name/description (RunningHub integration for image/video/audio generation) align with requiring a RUNNINGHUB_API_KEY and providing wrappers for apps like qwen-text-to-image. However the skill also requires/assumes a full RHMCP project to be cloned and built and references an OpenClaw MCP configuration (editing ~/.openclaw/openclaw.json). Asking the user to install a separate server project and to point the agent at it is heavier-weight than a simple API wrapper and expands the attack surface.
Instruction Scope
SKILL.md explicitly instructs cloning https://github.com/AIRix315/RHMCP, running npm install and npm run build, creating service.json and .env, and editing OpenClaw config. The included executor.mjs reads local files (service.json, apps.json, recommended-apps.json), and tries multiple directories (RHMCP_CONFIG, repo root, cwd). The runtime instructions therefore direct the agent/operator to download and run remote code and to read local config files beyond the declared env var. The skill's instructions also refer to RHMCP_CONFIG and file paths not declared in requires.env, which is a scope mismatch.
Install Mechanism
No formal install spec in registry, but README/SKILL.md tell users to git clone and run npm install/build on an external GitHub repo (RHMCP). That pulls and executes third-party code on the user's machine. Because this is an out-of-band install (not vetted by the registry), it increases risk and should be treated carefully.
Credentials
The declared required env var is RUNNINGHUB_API_KEY (appropriate for RunningHub). However the code and docs also reference RHMCP_CONFIG and expect a service.json/.env in the RHMCP directory; RHMCP_CONFIG is not declared in requires.env. The executor reads local files (service.json, apps.json) from multiple directories, which could cause the skill to consume or expose unrelated local configuration if RHMCP_CONFIG is pointed broadly. Overall the number and nature of config/env accesses are larger than declared.
Persistence & Privilege
always is false and disable-model-invocation is default (agent may call it autonomously). The skill's docs instruct manual edits to ~/.openclaw/openclaw.json to register the MCP server — that requires user action but would modify agent configuration if followed. The skill itself does not request permanent elevated privileges in metadata, but its installation instructions encourage changing agent config and running a long-lived local service, which increases persistence and attack surface if misused.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install runninghub-skill - After installation, invoke the skill by name or use
/runninghub-skill - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of rhmcp-skill (RunningHub AI integration):
- Provides tools to generate images, videos, and audio content via RunningHub AI.
- Agent decision logic for common user requests like text-to-image, image-to-image, and app listing.
- Async job handling with polling strategies for long tasks.
- Outlines parameter mapping, error codes, and failover strategies.
- Includes example workflows, storage guidelines, recommended templates, and troubleshooting tips.
Metadata
Frequently Asked Questions
What is rh-skill?
RunningHub AI 智能调用。Use when user wants to generate images, videos, or audio content. It is an AI Agent Skill for Claude Code / OpenClaw, with 132 downloads so far.
How do I install rh-skill?
Run "/install runninghub-skill" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is rh-skill free?
Yes, rh-skill is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does rh-skill support?
rh-skill is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created rh-skill?
It is built and maintained by AIRix315 (@airix315); the current version is v1.0.0.
More Skills