810872915-web/sally-script-breakdown-skill

Video Script Breakdown

Deploy the bundled single-file workbench instead of rebuilding its interface from scratch. The fixed asset preserves the interaction model that has already been refined through user testing.

What this skill provides

The workbench supports this analysis chain:

文字意思 → 原始文稿 → 文字功能 → 对应画面 → 画面表达

• Blue text: content meaning only.
• Yellow text: writing function only.
• Green text: the same passage carries both meaning and function.
• Left and right annotation cards omit copied source text to stay compact.
• A function annotation can link to one or more video frames.
• Uploaded frames have no surrounding card, border, title bar, timestamp, or footer.
• Drag any visible part of a frame to reposition it.
• Right-click a frame to replace it, focus its visual analysis, or delete it.
• All processing is local in the browser. Images are not uploaded anywhere.

Default action: deploy the tool

When the user asks to open, use, create, or share the tool:

1. Resolve this skill directory from the location of this SKILL.md.

2. Choose a user-facing destination inside the current workspace, normally:

   <workspace>/outputs/video-script-breakdown/

3. Run:

   python3 <skill-directory>/scripts/deploy.py \
     --output <workspace>/outputs/video-script-breakdown
   

4. Read the JSON printed by the script. Use its path value as the final deliverable.

5. Verify the deployed file exists and is non-empty.

6. Return a clickable absolute local link to index.html and one short usage sentence.

Do not start a server. The deployed HTML is designed to open directly with file://.

How to guide the user

Keep the handoff short:

1. Click “粘贴文稿” and paste the full transcript.
2. Select text and choose “标为意思” or “标为功能.”
3. Use “关联画面” under a function annotation.
4. Right-click the frameless placeholder to add a screenshot.
5. Drag the image itself to reposition it.
6. Write the visual meaning in the far-right blue panel.

Tell the user that the current standalone edition keeps state only during the open browser session. Refreshing or closing the page resets the draft. Do not claim that it has persistence, cloud sync, video-frame extraction, or AI analysis.

Handling supplied content

If the user supplies a transcript, deploy the tool and tell them to paste it using the in-page dialog. Do not silently rewrite the bundled asset with their private content.

If the user supplies screenshots, keep them local. The tool lets the user select or drag those files from their own computer. Do not upload images to external services unless the user explicitly requests that separate action.

If the user asks for automatic analysis, explain that this edition is an annotation workbench rather than an AI classifier. You may help draft meaning/function/visual notes in chat, but do not pretend the HTML performs automatic inference.

Customizing the workbench

When the user requests interface changes:

1. Deploy a copy first.
2. Edit the deployed copy, not assets/video-script-breakdown.html, unless the user explicitly asks to update the reusable skill itself.
3. Preserve the five-part relationship and the color semantics unless the request changes them.
4. Keep uploaded images frameless and directly draggable.
5. Keep image modification actions in the image context menu.
6. Run a browser smoke test after changes: load page, paste transcript, create an annotation, create a frame, load an image, drag it, open its context menu, and confirm no page errors.

Bundled resources

• assets/video-script-breakdown.html — complete offline workbench.
• scripts/deploy.py — dependency-free, non-destructive deployment helper.

Expected final response

Use this compact structure:

图文拉片工具已经准备好：

[打开工具](/absolute/path/to/index.html)

点击“粘贴文稿”开始；图片在“关联画面”后通过右键添加。