version: 0.0.1
name: describe
about: Describe a screenshot from clipboard for text-only LLMs (Cmd+Shift+Ctrl+4, Ctrl+Print)
notify: true
api:
output: dialog
input: query|image
actions:
- tag: tag_raw
run: vision
expect: string
action: |
{query|image|load|clipboard:screenshot|text}
[Task]
Describe this screenshot in rich detail.
This is a screenshot of an application interface, not a code extraction task.
Include:
- All visible text (if any)
- Layout, UI elements, objects, people, animals
- What is happening in the screenshot
- Colors: list each distinct color with its hex code, and what it is used for
Be thorough.
- tag: tag_md
run: small
expect: string
action: |
[Task]
Format the following description into clean Markdown.
- Use headers, lists, bold, and code blocks where appropriate
- Preserve ALL content — do not remove or summarize anything
- Keep the original language and wording
- Output ONLY the formatted Markdown
[Input]
{tag_raw}
- tag: tag_clipboard
run: value
expect: string
action: '{tag_md|clipboard}'