rh-skill

RunningHub AI 智能调用。Use when user wants to generate images, videos, or audio content.

Security Scan

The skill mostly matches its stated image/video/audio-generation purpose, but it instructs installing and running an external RHMCP project and its CLI which introduces filesystem access and extra environment/config requirements that aren't fully declared — this mismatch raises caution.

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.

What to consider before installing

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.

RHMCP Skill - OpenClaw 专用

概述

RHMCP 的 OpenClaw Skill 包装层，提供智能决策和 Agent 指引。

前置条件

1. 安装 RHMCP：git clone https://github.com/AIRix315/RHMCP.git && cd RHMCP && npm install && npm run build
2. 配置 OpenClaw MCP（见 README.md）
3. 创建 RHMCP 配置：service.json 和 .env

---

Agent 决策树

场景 → APP 映射

用户请求
    │
    ├─ "生成图片" / "文生图" / "画画"
    │       └─► rh_execute_app({ alias: "qwen-text-to-image", params: { text: "描述" } })
    │
    ├─ "修改图片" / "改图" / "图生图"
    │       └─► rh_execute_app({ alias: "qwen-image-to-image", params: { image: "URL", prompt: "描述" } })
    │
    └─ "更多APP"
            └─► rh_list_apps() 查看所有可用APP

参数映射

文生图 (qwen-text-to-image)

参数	必填	默认值	说明
text	✅	-	提示词
width	❌	1024	图片宽度
height	❌	1024	图片高度

图生图 (qwen-image-to-image)

参数	必填	默认值	说明
image	✅	-	输入图片URL
prompt	✅	-	修改提示词

---

轮询策略

长时间任务使用异步模式：

APP 类型	首次等待	轮询间隔	最大等待
图生视频	3 分钟	1 分钟	15 分钟
数字人	2 分钟	30 秒	10 分钟
文生图	10 秒	5 秒	5 分钟

// 提交异步任务
const { taskId } = await rh_execute_app({
  alias: "image-to-video",
  params: { image: "https://..." },
  mode: "async",
});

// 轮询查询
await sleep(180000); // 3分钟
while (true) {
  const status = await rh_query_task({ taskId });
  if (status.status === "SUCCESS") return status.data;
  if (status.status === "FAILED") throw new Error(status.msg);
  await sleep(60000); // 1分钟
}

---

存储策略

步骤位置	存储模式	原因
中间步骤	none	URL 直接传递
最终输出	none	返回 URL，用户保存
用户要求	network	用户明确要求保存

// 中间步骤：URL 直接传递
const img = await rh_execute_app({
  alias: "qwen-text-to-image",
  params: { text: "一只猫" },
});

// 后续步骤：使用 URL
const video = await rh_execute_app({
  alias: "image-to-video",
  params: { image: img.outputs[0].originalUrl, motion: "转身" },
});

---

链式工作流

// 步骤1：文生图
const img = await rh_execute_app({
  alias: "qwen-text-to-image",
  params: { text: "一个女孩在樱花树下" },
});

// 步骤2：图生视频
const video = await rh_execute_app({
  alias: "image-to-video",
  params: {
    image: img.outputs[0].originalUrl,
    motion: "樱花飘落，女孩转身微笑",
  },
});

return { videoUrl: video.outputs[0].originalUrl };

---

失败重试

const fallbacks = ["sdxl-text-to-image", "flux-text-to-image"];
for (const alias of fallbacks) {
  try {
    return await rh_execute_app({ alias, params });
  } catch (e) {
    continue; // 尝试下一个
  }
}
throw new Error("所有备选 APP 均失败");

---

错误码速查

错误码	含义	处理方式
0	成功	正常返回结果
804	排队中	等待后轮询
805	任务失败	调整提示词重试
813	执行中	继续等待
421	并发限制	等待 5 秒后重试

---

工具清单

工具	用途
rh_list_apps	列出所有可用APP
rh_get_app_info	获取APP参数详情
rh_execute_app	执行APP任务
rh_query_task	查询任务状态
rh_upload_media	上传媒体文件

---

推荐模板

模板ID	名称	尺寸	适用场景
anime-portrait	动漫人像	1024×1024	角色生成
landscape-wide	风景横图	1920×1080	风景创作
portrait-vertical	人像竖图	768×1024	人物肖像

// 使用模板
const templates = require("./references/templates.json");
const t = templates.templates["anime-portrait"];
rh_execute_app({ alias: t.alias, params: { ...t.params, text: "用户描述" } });

---

故障排除

问题	原因	解决方案
"未找到 alias"	APP 未配置	运行 rhmcp --update-apps
"apiKey 是必填项"	未配置 API Key	在 .env 设置 RUNNINGHUB_API_KEY
"任务超时"	执行时间过长	使用 mode: "async"

---

License

MIT-0