功能描述

高密度信息大图生成技能。根据文章/内容生成高信息密度的视觉大图，适合知识干货、笔记分享、课程内容可视化。支持9种风格：坐标蓝图波普实验室、复古波普网格、文件夹风、色块热敏纸、复古手帐、档案混合媒介、色块酸性风、票据剧场戏票风、Claude陶土风。使用系统配置的图像生成模型（建议4K）生成。当用户提供文章/内容并要...

使用说明 (SKILL.md)

高密度信息大图（hd-infoimage）

Name: Hd Infoimage
Author: dadaniya99

根据用户提供的文章或内容，生成高信息密度的视觉大图。

使用方式

用户提供文章内容后，告知龙虾：

根据给你的文章和这个 prompt，生成适合数量的高信息密度大图

风格选项（9种）

从 references/ 中选择对应风格文件，将提示词 + 用户文章内容一起发给图像模型。

#	风格名称	特征	文件
1	坐标蓝图·波普实验室	实验室精密感，视觉坐标系，高密度数据	`style-01-坐标蓝图·波普实验室.md`
2	复古波普网格风	70年代复古波普，Swiss Grid，6-7模块	`style-02-复古波普网格风.md`
3	文件夹·热敏纸风	文件夹质感，打印热敏纸，英文版	`style-03-文件夹风格（打印热敏.md`
4	色块·热敏纸风	高对比色块框架+中性纸张，英文详细版	`style-04-色块·热敏纸风（英文.md`
5	复古手帐·档案风	侦探证据板美学，手帐剪贴风	`style-05-复古手帐·档案风.md`
6	档案·混合媒介风	牛皮纸+深黑，证书徽章，英文版	`style-06-档案·混合媒介风（英.md`
7	色块·复古未来酸性风	酸性涂鸦，复古未来主义，高对比	`style-07-色块·复古未来酸性风.md`
8	票据·剧场戏票风	五幕剧叙事，剧场票据，粒状纹理	`style-08-票据·剧场戏票风.md`
9	Claude陶土风 ⭐	温暖陶土橙+炭黑+米白，手绘涂鸦，粗线条描边，几何抽象，扁平化，亲和力与专业感并存	`style-09-Claude陶土风.md`

未指定风格时，默认推荐风格2（复古波普网格）或风格1（坐标蓝图）。

生成图片

使用系统配置的图像生成模型生成图片。

# 通过系统工具或配置的图像生成服务生成
# 输出到 /root/myfiles/\x3Cfilename>.png
# 建议4K质量

注意事项

复古风提示词里如出现"小红书" logo，可删掉提示词中的"小红书"字样
每张图建议包含 6-7 个子主题模块，信息密度要高
生成后用 send_to_feishu.sh 发送（见 sketch-illustration skill）

安全使用建议

Key things to consider before installing or enabling this skill: - The README contains an explicit method to extract a ZenMux API key from the agent's config file (~/.openclaw/openclaw.json). The skill metadata does NOT declare any required credentials — ask the author to stop recommending reading agent config files and to declare required credentials (e.g., ZENMUX_API_KEY) explicitly instead. - If the skill needs to call an external image provider, have the author provide a clear, minimal set of environment variables (and document their scope) rather than instructing the agent to read internal config files. - Confirm where generated images are stored and transmitted: the README uses /root/myfiles and references a send_to_feishu.sh helper. Verify what send_to_feishu.sh does and whether it posts images to external services; require explicit consent before any automatic upload. - Prefer least privilege: if you must provide an API key, provision a dedicated key with narrow scope and rotate/delete it when not in use; do not store long-lived privileged keys in agent config accessible to all skills. - Ask the author to remove any instructions that access other skills' workspaces or the agent's global config; instead, make dependencies explicit (e.g., 'this skill requires the zenmux-image-generation skill and a ZENMUX_API_KEY set in the skill environment'). - If you lack trust/confidence, run the skill in a sandboxed environment (no access to ~/.openclaw, limited outbound network) or decline until the author clarifies these inconsistencies. If the author can (1) remove the cat ~/.openclaw/... command, (2) declare required env vars and justify them, and (3) document any external uploads, the skill would be much more coherent and lower-risk.

功能分析

Type: OpenClaw Skill Name: hd-infoimage Version: 1.2.0 The skill provides templates and instructions for generating high-density infographics. It is classified as suspicious because the README.md explicitly instructs the agent to execute a shell command that parses the core OpenClaw configuration file (~/.openclaw/openclaw.json) to extract a ZenMux API key. While this is functionally necessary for the stated purpose of image generation, teaching an AI agent to programmatically scrape its own configuration for secrets significantly increases the risk of credential theft if the agent is targeted by prompt injection through the user-provided article content.

能力标签

requires-sensitive-credentials

能力评估

⚠ Purpose & Capability

The declared purpose is to generate high-density infographics using a system-configured image model — that is coherent. However, the README explicitly references the ZenMux/Imagen 3 provider and shows a workflow that requires a ZENMUX_API_KEY and a path under /root/.openclaw/workspace/skills/zenmux-image-generation. The skill's metadata declares no required credentials or config paths, so asking for a provider API key (and pointing to another skill's workspace) is disproportionate and inconsistent with the stated requirements.

⚠ Instruction Scope

SKILL.md is instruction-only but the README contains concrete commands that reference local files and configuration: it shows running a python script in another skill's workspace, writing output to /root/myfiles/<filename>.png, and (critically) reads an API key via `cat ~/.openclaw/openclaw.json | python3 -c '... print(d["models"]["providers"]["ZenMux"]["apiKey"])'`. Those instructions tell the agent to read agent-local secrets and to access other skill directories — behavior outside the narrow remit of 'generate an image from user content' and not declared in requires.env.

✓ Install Mechanism

There is no install spec and no code files to execute from the skill itself (instruction-only), which is lower risk. That said, the README assumes the presence of another skill's workspace and Python scripts (zenmux-image-generation) — relying on external workspaces increases operational coupling but is not an installer risk in itself.

⚠ Credentials

The skill declares no required environment variables or credentials, yet the README/README snippet expects a ZENMUX_API_KEY and demonstrates extracting it from the agent's ~/.openclaw/openclaw.json. Requesting access to a system-stored API key (without declaring it) is disproportionate and a secret-exposure risk. The skill also suggests writing to /root/myfiles and calling send_to_feishu.sh (implying external transmission), which should have been explicitly declared and justified.

ℹ Persistence & Privilege

always:false and no claimed modification of agent configuration — good. However, the instructions reference other skills' tools (zenmux-image-generation, sketch-illustration/send_to_feishu.sh) and a path under the agent workspace, which enables cross-skill access. Autonomous invocation isn't disabled; combined with the secret-access instructions this raises the blast radius if the agent runs the skill without user oversight.

版本历史

v1.2.0

高密度信息大图生成技能，支持9种风格，使用系统配置图像模型

v1.1.0

- 新增“Claude陶土风”视觉风格，现共支持9种高密度信息大图风格 - 更新文档说明与风格表，添加陶土风描述与指引 - 新增 references/style-09-Claude陶土风.md 参考文件 - 增加 README.md 文件，优化技能使用说明

v1.0.0

首发：根据文章内容生成高密度信息大图，支持8种风格（坐标蓝图波普/复古波普网格/文件夹/热敏纸/手帐/档案/酸性/票据），使用Imagen 3生成

元数据

Slug hd-infoimage

版本 1.2.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 3

常见问题

Hd Infoimage 是什么？

高密度信息大图生成技能。根据文章/内容生成高信息密度的视觉大图，适合知识干货、笔记分享、课程内容可视化。支持9种风格：坐标蓝图波普实验室、复古波普网格、文件夹风、色块热敏纸、复古手帐、档案混合媒介、色块酸性风、票据剧场戏票风、Claude陶土风。使用系统配置的图像生成模型（建议4K）生成。当用户提供文章/内容并要... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 507 次。

如何安装 Hd Infoimage？

在 OpenClaw 或 Claude Code 对话框中运行命令「/install hd-infoimage」即可一键安装，无需额外配置。

Hd Infoimage 是免费的吗？

是的，Hd Infoimage 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。

Hd Infoimage 支持哪些平台？

Hd Infoimage 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 Hd Infoimage？

由 dadaniya99（@dadaniya99）开发并维护，当前版本 v1.2.0。

Hd Infoimage