← 返回 Skills 市场
Sketch Illustration
作者
dadaniya99
· GitHub ↗
· v1.2.0
· MIT-0
675
总下载
0
收藏
0
当前安装
4
版本数
在 OpenClaw 中安装
/install sketch-illustration
功能描述
插画图片生成技能,支持多种手绘风格。适合流程图、功能说明、PPT配图、教程配图、知识图和手绘信息图等场景。支持五种风格:A) Sketch 极简手绘风(Notion/Linear 风格,简笔人物,冷淡低饱和);B) Watercolor 奶油彩铅水彩风(暖色调,纸纹彩铅,适合PPT讲义);C) Flat Vect...
安全使用建议
Before installing or enabling this skill, consider the following:
- The skill package contains runnable scripts that will read /root/.openclaw/openclaw.json and may use ZENMUX_API_KEY and Feishu app credentials to call external APIs. The skill metadata did not declare these credential requirements — expect it to access your agent's stored keys.
- It will write generated images under /root/myfiles/ and call a Feishu send script which will post images to a hardcoded open_id (ou_22f2eefd5abe63e0cd67f4882cec06d4) unless you change usage. If you don't want automatic uploading, do not run send_to_feishu.sh and remove or inspect its calls first.
- Recommended mitigations: review the contents of ~/.openclaw/openclaw.json to see what keys would be exposed; run the skill in a sandboxed agent or with limited/test credentials; require the skill to accept explicit API tokens via environment variables (and document them) rather than reading global agent config; inspect and, if needed, remove or modify scripts that call Feishu.
- Additional notes: there are some implementation inconsistencies (e.g., different capitalizations for ZenMux keys, fixed /root paths, and a prompt example referencing 'scraping Reddit/X' in a demonstration image) — these could be bugs or sloppy engineering. If you need to use this skill, ask the publisher to: (1) declare required credentials in the manifest, (2) avoid reading global agent config without explicit consent, and (3) remove hardcoded recipient IDs or require user confirmation before sending.
功能分析
Type: OpenClaw Skill
Name: sketch-illustration
Version: 1.2.0
The skill bundle contains scripts that extract sensitive credentials (API keys and Feishu App Secrets) from the global OpenClaw configuration file (`/root/.openclaw/openclaw.json`). While these are used for the stated purpose of image generation and delivery, the skill hardcodes a specific Feishu recipient ID (`ou_22f2eefd5abe63e0cd67f4882cec06d4`) in both `SKILL.md` and `scripts/send_to_feishu.sh`. This configuration causes all generated content to be automatically exfiltrated to a specific external user, which poses a significant data leakage risk if the agent is used to visualize sensitive or proprietary information.
能力标签
能力评估
Purpose & Capability
The declared purpose is image/illustration generation, which normally needs only an image-model API key. However, the included scripts attempt to read /root/.openclaw/openclaw.json and environment variables to obtain ZenMux (image) API keys and Feishu channel credentials, and call another skill's Feishu send script. The skill metadata listed no required env vars or credentials, so the runtime behavior is not reflected in the manifest.
Instruction Scope
SKILL.md instructs generating images to /root/myfiles and running scripts/send_to_feishu.sh to upload to Feishu (with a hardcoded open_id). The scripts read local agent config files (~/.openclaw/openclaw.json) for sensitive tokens and call external APIs. That expands scope beyond 'generate only' to reading internal config and sending data to external endpoints.
Install Mechanism
There is no install spec (instruction-only), which minimizes disk-write/install risk. The presence of scripts increases runtime actions but nothing in the manifest installs third-party binaries or downloads remote archives.
Credentials
No env vars or credentials are declared, yet scripts expect ZENMUX_API_KEY (env) or to read openclaw.json for ZenMux keys and Feishu appId/appSecret. Requesting access to the agent's central config file (openclaw.json) is disproportionate to the stated purpose unless explicitly declared — it gives the skill access to multiple provider credentials and channel secrets.
Persistence & Privilege
The skill is not marked always:true, and autonomous invocation is allowed by default. Still, the scripts read the agent's configuration and call another skill's Feishu send script, which effectively uses platform credentials at runtime. That coupling to internal agent config increases blast radius if the skill acts without user confirmation.
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 sketch-illustration - 安装完成后,直接呼叫该 Skill 的名称或使用
/sketch-illustration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
v1.2.0
Update to 1.2.0
v1.1.1
移除描述中的具体 API 提及(Imagen 3/ZenMux),改为通用描述
v1.1.0
Add doodle infographic templates for mechanism diagrams, comparison layouts, and step-by-step flow charts; expand style guidance and keep Feishu image sending on the stable upload path.
v1.0.0
首发:支持3种手绘插画风格(Sketch极简/Watercolor彩铅/Flat Vector复古矢量),含5步配图工作流和布局模板,生成后自动发送到飞书
元数据
常见问题
Sketch Illustration 是什么?
插画图片生成技能,支持多种手绘风格。适合流程图、功能说明、PPT配图、教程配图、知识图和手绘信息图等场景。支持五种风格:A) Sketch 极简手绘风(Notion/Linear 风格,简笔人物,冷淡低饱和);B) Watercolor 奶油彩铅水彩风(暖色调,纸纹彩铅,适合PPT讲义);C) Flat Vect... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 675 次。
如何安装 Sketch Illustration?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install sketch-illustration」即可一键安装,无需额外配置。
Sketch Illustration 是免费的吗?
是的,Sketch Illustration 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。
Sketch Illustration 支持哪些平台?
Sketch Illustration 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 Sketch Illustration?
由 dadaniya99(@dadaniya99)开发并维护,当前版本 v1.2.0。
推荐 Skills