MiniMax Coding Plan Tool

A lightweight MiniMax Coding Plan Tool skill that directly calls official MCP APIs using pure JavaScript. No external MCP servers. No tool bridge. No subproc...

Security Scan

The skill's code, runtime instructions, and required credential are consistent with a simple MiniMax web-search and image-understanding tool; nothing appears to be covertly exfiltrating unrelated secrets, though there are a few small documentation inconsistencies and the normal privacy risk of uploading local images to a remote API.

Purpose & Capability

The declared purpose (web search + image understanding) matches the code and required MINIMAX_API_KEY. Minor inconsistencies in documentation exist (subscribe URL uses domain 'platform.minimaxi.com' while the script calls 'api.minimax.chat'; some CLI names differ), but these look like documentation/typo issues rather than functional mismatches.

Instruction Scope

SKILL.md instructs the agent to run the provided Node script and to provide an API key; the script only reads local image files when explicitly asked and uploads them (base64) to the remote MiniMax API. The README warns about local image leakage. The instructions do not direct reading of unrelated files or other environment variables.

Install Mechanism

No install spec (instruction-only plus a small JS file). The skill requires 'node' to run the included script; no downloads or archive extraction are performed by the install process.

Credentials

Only a single credential (MINIMAX_API_KEY) is required and it is used as the bearer token for the API calls. No unrelated credentials or config paths are requested.

Persistence & Privilege

Skill is not always-enabled, does not modify other skills or system-wide settings, and does not request persistent elevated privileges.

Assessment

This skill appears to do what it says: it makes HTTPS calls to api.minimax.chat and will upload local images (converted to base64) to that service. Before installing: 1) Verify you trust the MiniMax endpoint (api.minimax.chat) and that the subscribe URL mismatch (platform.minimaxi.com) is just a typo and not a malicious redirect. 2) Only use a Coding Plan API key as required — do not give unrelated credentials. 3) Do not submit sensitive or regulated images because local files will be uploaded to the remote API (the SKILL.md warns this). 4) Prefer storing the key in a secure secret store and verify the script file location and contents in your environment before running. If you want extra assurance, ask the publisher to correct the doc inconsistencies (domains and CLI name) and confirm the official API endpoint.

⚠️ Requires a Coding Plan API Key Subscribe at: https://platform.minimaxi.com/subscribe/coding-plan

A lightweight MiniMax Coding Plan Tool skill that directly calls official MCP APIs using pure JavaScript. No external MCP servers. No tool bridge. No subprocess invocation.

Designed for minimal overhead and fast integration within OpenClaw.

---

✨ Features

This skill provides two native capabilities of MiniMax Coding Plan MCP:

1. minimax_web_search

Web search powered by MiniMax Coding Plan API.

• Real-time information retrieval
• Lightweight API-based implementation
• No external search engine dependency

2. minimax_understand_image

Vision-language image understanding via MiniMax VLM.

• Image semantic analysis
• Visual content description
• Multimodal reasoning support

---

🧩 Architecture

• Pure JavaScript implementation
• Direct HTTPS API calls
• No MCP server runtime required
• No external tool dependency
• Minimal resource usage

---

🔑 Configuration

openclaw config set skills.entries.minimax-coding-plan-tool.apiKey "sk-your-key"

or

{
  "skills": {
    "entries": {
      "minimax-coding-plan-tool": {
        "apiKey": "sk-your-key"
      }
    }
  }
}

Read the API key from openclaw.json.

---

Tool Discovery

Execute minimax_coding_plan_tool.js with environment variable MINIMAX_API_KEY to dynamically register these tools:

MINIMAX_API_KEY="sk-your-key" node minimax_coding_plan_tool.js tools

Output format:

{
  "tools": [
    {
      "name": "minimax_web_search",
      "description": "...",
      "inputSchema": { ... }
    },
    {
      "name": "minimax_understand_image",
      "description": "...",
      "inputSchema": { ... }
    }
  ]
}

---

Tool 1 — minimax_web_search

Purpose

Real-time web search using MiniMax Coding Plan search API.

CLI Invocation

MINIMAX_API_KEY="sk-your-key" node minimax_coding_plan_tool.js web_search "<query>"

Example:

MINIMAX_API_KEY="sk-your-key" node minimax_coding_plan_tool.js web_search "OpenAI GPT-5 release date"

Input Schema

{
  "query": "string"
}

Output Format

Success:

{
  "success": true,
  "query": "...",
  "results": [
    {
      "title": "...",
      "link": "...",
      "snippet": "...",
      "date": "..."
    }
  ],
  "related_searches": []
}

Error:

{
  "error": "error message"
}

---

Tool 2 — minimax_understand_image

Purpose

Image understanding using MiniMax Coding Plan VLM API. Only mainstream image formats are supported (for example: JPEG/JPG, PNG, WebP, and GIF).

Supports:

• HTTP / HTTPS image URLs
• Local file paths
• @localfile.jpg shorthand

Local files are automatically converted to base64 data URLs.

Security Notice

This tool requires outbound network access. If a local image is provided, its content is transmitted to the remote MiniMax API for processing, which introduces a potential risk of local image data leakage. Do not submit sensitive, private, or regulated images unless you fully accept this risk.

CLI Invocation

MINIMAX_API_KEY="sk-your-key" node minimax_coding_plan_tool.js understand_image <image_source> "<prompt>"

Examples:

Remote image:

MINIMAX_API_KEY="sk-your-key" node minimax_coding_plan_tool.js understand_image https://example.com/image.jpg "Describe this image"

Local image:

MINIMAX_API_KEY="sk-your-key" node minimax_coding_plan_tool.js understand_image ./photo.png "What objects are visible?"

With @ prefix:

MINIMAX_API_KEY="sk-your-key" node minimax_coding_plan_tool.js understand_image @photo.png "Summarize the scene"

Input Schema

{
  "image_source": "string",
  "prompt": "string"
}

Output Format

Success:

{
  "success": true,
  "prompt": "...",
  "image_source": "...",
  "analysis": "model response"
}

Error:

{
  "error": "error message"
}