Python Screenshot Assistant

Problem

In Agent conversations, screenshot requests are rarely just “save the full screen.” Users may want to capture a specific window, scroll through a long page, annotate key areas, or blur sensitive information. Manual workflows usually require switching between capture, clipboard, and editing tools while parameters and paths remain scattered.

How It Works

@user_3cdbf8ea/screenshot-assistant runs as a Hermes Agent integration tool and exposes Python functions in scripts/tool.py:
- capture_screen(path): captures the full screen and saves it.
- capture_window(title, path): captures a window by fuzzy title match.
- Capture modes: region, window, full screen, and scroll capture.
- Annotation and editing: arrows, rectangles, circles, lines, and text.
- Privacy handling: applies mosaic or blur to selected areas.
- Color picking: reads pixel colors from the screen.
- Output: saves PNG/JPG or copies the image to the clipboard.
It prefers PIL(Pillow) ImageGrab and falls back to Windows ctypes when needed. Functions return standardized JSON with path and file size, and handle errors such as missing window matches.

Boundaries

This skill fits screenshot capture, annotation, color picking, and lightweight editing. For complex image recognition, cross-platform GUI automation, or high-volume batch capture, validate separately. Provide a clear window title, output format, and annotation parameters to reduce fuzzy-match misfires.