Snap Illustrator

Instantly analyze markdown articles, generate prompts, and seamlessly insert AI illustrations with zero configuration required. Supports Pollinations API (zero-config) with graceful fallback to HuggingFace.

Instantly analyze long-form markdown articles, extract core scenes, generate prompts, and seamlessly insert AI illustrations with a snap. Designed for blogs, articles, and documentation with zero configuration required.

Trigger Scenarios

When the user asks to "illustrate this article", "add images to this markdown", "为文章配图", or "加配图".

Design Philosophy & Zero-Config Flow

Default to Zero-Config Experience:

1. By default, utilize the included Node.js script to run via the unauthenticated Pollinations.ai API. This process is completely zero-config and requires NO API tokens from the user.
2. As long as images generate successfully, DO NOT ask the user to configure any tokens. Let the user experience an instant, frictionless image insertion.
3. ONLY IF the Node.js script throws an exception or logs PLEASE_ASK_USER_FOR_TOKEN, indicating the free unauthenticated service is rate-limited or unavailable, should you proactively ask the user: "The free generation service is currently busy. Do you have a HuggingFace Token you could provide to continue?"
4. If the user provides a HuggingFace Token, instruct them to set it as an environment variable (e.g., export HF_TOKEN="<your_token>" in their ~/.zshrc or ~/.bash_profile), or temporarily use the token to run the script. AI agents should NOT permanently save tokens to disk unless explicitly requested by the user.

Execution Workflow

Step 1: Text Analysis & Planning

1. Read Content: Read the user's target Markdown file.
2. Segmentation: Analyze the structural and semantic boundaries of the text and partition it into sections suitable for illustrations (e.g., placing one image per major heading or roughly one per 300 words).
3. Propose an Outline: Before spending time generating images, present a brief "Illustration Outline" to the user. Describe how many images you plan to insert, where they will be placed, and what concepts they represent. Wait for the user's confirmation before proceeding.

Step 2: Generate Prompts Once the user approves the outline, design high-quality, English-only image generation prompts for each selected insertion point.

• Suggested format: [Core Concept/Subject], [Setting/Background], [Style Keywords], 4k, hyper-detailed, masterpiece
• Avoid Text Generation Constraints: Explicitly design prompts to avoid rendering text, letters, or numbers to prevent garbled artifacts. Append a uniform style keyword (e.g., flat vector illustration) unless the user specified a distinct style.

Step 3: Execute Node.js Script Invoke the included Node.js script using run_in_terminal to perform the actual image generation.

Locating the script: This SKILL.md file is your reference — scripts/generate.mjs lives in the same directory as this file. You already know the absolute path to this SKILL.md (it was provided to you when the skill was loaded). Strip the filename to get the skill root directory, then append scripts/generate.mjs.

For example, if this skill file is at /path/to/skills/snap-illustrator/SKILL.md, then run:

bash

node /path/to/skills/snap-illustrator/scripts/generate.mjs --prompt "Your English Prompt Here" --output "<workspace>/images/img_name.jpg"

(Note: Ensure <workspace> points to the absolute path of the markdown file's directory. Images should typically be placed in an images/ subdirectory adjacent to the markdown file)

Step 4: Finalize & Insert Insert standard Markdown image syntax referencing the generated images into the original text at the appropriate boundaries:

markdown

## Heading

![Scene: Description...](images/img_name.jpg)
Body paragraph...

Notify the user once all illustrations have been successfully inserted.

Troubleshooting & Connectivity Issues

Handle failures in the following priority order:

Issue 1: Unauthenticated Service Unavailable

If the Node.js script fails immediately (non-network error), exits with a non-zero code, or outputs PLEASE_ASK_USER_FOR_TOKEN:

Immediately tell the user:

"免费的无认证图片生成服务当前不可用（速率限制或服务故障）。建议切换到 HuggingFace API。请问您是否有 HuggingFace Token？"

Do not retry the free service. Proceed to Issue 2 for HuggingFace setup.

Issue 2: Network Connectivity Problems (especially Mainland China)

If the script times out, throws a network/fetch error, or ECONNREFUSED/ENOTFOUND/ETIMEDOUT appears in the output, the user may be experiencing connectivity issues with overseas services.

Ask the user:

"检测到网络连接问题。请确认：

1. 您是否能正常访问 https://huggingface.co ？
2. 您是否需要配置 HTTP/HTTPS 代理来访问境外服务？"

If the user cannot access HuggingFace or needs a proxy, proceed to Issue 3.

Issue 3: HTTP Proxy Configuration

If the user confirms they have an HTTP proxy, ask them to provide the proxy address:

"请提供您的 HTTP 代理地址（格式如 http://127.0.0.1:7890）："

Once the user provides the proxy address, run the script with proxy environment variables:

bash

HTTPS_PROXY="<proxy_url>" HTTP_PROXY="<proxy_url>" node /path/to/skills/snap-illustrator/scripts/generate.mjs --prompt "..." --output "..."

Recording the proxy for future use: Update this SKILL.md file by appending the user's proxy setting to the frontmatter metadata section, so it can be reused automatically in subsequent runs:

yaml

metadata:
  openclaw:
    requires:
      env:
        - HF_TOKEN
      bins:
        - node
    http_proxy: "http://127.0.0.1:7890"   # ← add proxy here

And in all subsequent script executions within this session, automatically prepend HTTPS_PROXY="<proxy_url>" HTTP_PROXY="<proxy_url>" to the command without asking the user again.

Issue 4: All Services Unavailable

If none of the above resolves the issue, inform the user:

"图片生成服务当前无法访问。您可以稍后重试，或手动准备图片后，我可以帮您将图片链接插入到文章中。"