← 返回 Skills 市场

Drawthings Agent

Name: Drawthings Agent
Author: mijuu

作者 mijuu · GitHub ↗ · v0.1.1 · MIT-0

cross-platform ⚠ suspicious

141

总下载

当前安装

版本数

在 OpenClaw 中安装

/install drawthings-agent

功能描述

Generate images using Draw Things app via dt-skill CLI or MCP

使用说明 (SKILL.md)

Skill: Draw Things for AI Agents

Slug: drawthings-agent
Display Name: Draw Things for AI Agents

🛠 Setup & Verification (AI ACTION REQUIRED)

If dt-skill is not found or fails with "module not found", you MUST perform these steps:

Global Installation: Run npm install -g @mijuu/drawthings to ensure the CLI and MCP entry points are available.
Verify Environment: Run dt-skill doctor. If paths are missing, ask the user for their Draw Things model and server binary paths, then run dt-skill config --server-path "..." --models-path "...".

🧩 Protocol Preference (CRITICAL for AI)

This skill provides two interfaces. You MUST follow this priority:

Model Context Protocol (MCP): If you see the tool generate_image in your toolset, ALWAYS use it. It is the most reliable interface for AI.
CLI Interface (dt-skill): Use dt-skill gen ONLY if the MCP tool is missing or for server management (e.g., dt-skill server start).

🛠 Target Tool

MCP Tool: generate_image (Preferred)
CLI Command: dt-skill gen (Fallback)

🤖 AI Usage Instructions

1. Basic Generation

MCP: Call generate_image(prompt: "...").
CLI: Run dt-skill gen --prompt "...". You MUST provide a detailed prompt.

2. High Quality & Upscaling (CRITICAL)

If the user asks for "high quality", "high resolution", "4K", or "clear" images, you MUST use the --upscale parameter at the time of generation.

For 2x Upscale: --upscale 2
For 4x Upscale: --upscale 4 (Note: Takes significantly longer) Example: dt-skill gen --prompt "..." --upscale 2

3. Aspect Ratio & Resolution

Default is 1024x576 (16:9).

Square: --width 1024 --height 1024
Portrait: --width 576 --height 1024
Standard: --width 1024 --height 576

4. Model Selection

Default: z_image_turbo_1.0_q6p.ckpt (Fastest, 8 steps).
Listing Models: Use dt-skill models to see what is available before suggesting a change.

5. Task Control & Sessions

When running in environments that support background tasks (like OpenClaw or OpenCode), use is_background: true to allow real-time progress monitoring.

Progress: You will see Sampling step: X/Y.
Heartbeat: Every 30s you will see ... still working ....
Timeout: Default is 600s. For heavy 4x upscales, use --timeout 1200.

⚠️ CRITICAL CONSTRAINTS for AI

NO TRUNCATION: Never use "..." or "etc." in the prompt. Provide the complete text.
NO POST-PROCESSING: Upscaling MUST be part of the initial gen command.
RESPONSE COMPRESSION: Ensure "Response Compression" (FPY) is DISABLED in Draw Things settings.

📋 Parameter Reference (CLI)

Parameter	Type	Description
`--prompt`	String	(Required) Positive prompt
`--negative-prompt`	String	Elements to exclude
`--upscale`	Number	1, 2, or 4
`--model`	String	Filename of the model
`--steps`	Number	Number of sampling steps (8-30)
`--seed`	Number	0 for random, or specific number
`--output`	Path	Custom path for result
`--timeout`	Number	Max runtime in seconds (Default 600)

安全使用建议

This skill appears to do what it says (drive a Draw Things CLI or MCP), but take these precautions before following its setup steps: 1) Do not blindly run `npm install -g @mijuu/drawthings`. Verify the package on the npm registry (author, maintainers, version history) and inspect its source repository before installing. Prefer a local or isolated install (no -g) or run inside a sandbox/container if possible. 2) The skill will ask you for local model and server binary paths—only provide paths you trust; model files can be large and may contain sensitive or copyrighted material. 3) The SKILL.md tells you to change Draw Things settings (disable Response Compression FPY); understand what that setting does before changing it. 4) If you want lower risk, ask the skill author for a homepage, source repo, or signed package; if available and trusted, that would move this evaluation toward benign. 5) Because the skill suggests installing third-party code and changing system settings, treat it as potentially risky: review the package contents or run the tool in an isolated environment before giving the agent permission to perform these actions.

功能分析

Type: OpenClaw Skill Name: drawthings-agent Version: 0.1.1 The skill instructs the AI agent to perform a global system installation (`npm install -g @mijuu/drawthings`) and execute various shell commands for configuration and image generation. While these actions are aligned with the stated purpose of integrating the Draw Things app, the use of "MUST" directives to compel the agent to install software and the presence of a future-dated timestamp in _meta.json (March 2026) are unusual. The requirement for global installation and broad shell access constitutes a significant attack surface and potential supply chain risk.

能力评估

✓ Purpose & Capability

The name/description (generate images via Draw Things) matches the SKILL.md: it documents using an MCP tool `generate_image` or the `dt-skill` CLI and parameters for generation, models, upscaling, timeouts, etc. Requesting model and server binary paths is coherent for a local image-generation tool.

ℹ Instruction Scope

Instructions are specific and constrained to image generation (prompts, upscales, dimensions, model selection, timeouts). However, the skill mandates running `npm install -g @mijuu/drawthings` if the CLI is missing and asks the agent to request local file paths and to change Draw Things settings (disable FPY), which are system-level user actions beyond pure prompting.

⚠ Install Mechanism

There is no formal install spec, but the runtime instructions require a global npm install of `@mijuu/drawthings`. A global npm installation modifies the host environment and pulls code from the public registry; the skill metadata provides no homepage or source to verify the package. Installing packages globally without a verified source is a non-trivial risk.

✓ Credentials

The skill declares no environment variables, no credentials, and no config paths. Asking for model and server binary paths is consistent with operating a local model server and does not request unrelated secrets.

✓ Persistence & Privilege

The skill is not flagged with always:true, does not request permanent presence, and contains no instructions to modify other skills or system-wide agent config. The primary privilege concern is the suggested global npm install, which affects the system but is not an automatic persistence privilege.

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 drawthings-agent
安装完成后，直接呼叫该 Skill 的名称或使用 /drawthings-agent 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

v0.1.1

- Initial release of drawthings-agent skill. - Added support for generating images using Draw Things via both CLI and MCP interfaces. - Included detailed AI usage instructions and critical constraints for quality and protocol handling. - Provided setup, verification, and fallback steps for both MCP and CLI environments. - Documented all available parameters and model selection guidance.

v0.1.0

- Initial release of the Draw Things for AI Agents skill. - Provides both Model Context Protocol (preferred) and CLI interfaces for image generation. - Supports high quality, upscaling, aspect ratio/resolution adjustments, and model selection. - Includes task control options for background operation and real-time progress. - Details critical usage constraints and full parameter reference for consistent agent behavior.

元数据

Slug drawthings-agent

版本 0.1.1

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 2

常见问题