AI-ready descriptions

--describe prints a structured, plain-text summary of a completed capture run — ready to paste into a model alongside the screenshot files.

terminal


shotsweep capture --urls paths.json --describe


--- AI description ---

Screenshot batch summary
========================

Screenshots: 26
Pages: 3
Mode: full
Output: ./screenshots

Pages captured:

- https://example.com
  Screenshots: 17
  Viewports: 1440x900
  Capture: 17 section screenshots

- https://example.com/about
  Screenshots: 6
  Viewports: 1440x900, 390x844

- https://example.com/contact
  Screenshots: 3
  Viewports: 1440x900

AI review instructions:
Attach the screenshot files listed in manifest.json and review them for:
- visual hierarchy and layout
- responsive behavior across viewports
- spacing, typography, and consistency
- accessibility concerns
- obvious visual regressions or rendering issues

Compare related screenshots where multiple viewports or capture modes are available.

What this actually does

--describe does not analyze the screenshots itself — ShotSweep doesn’t ship a vision model. It produces a structured, AI-ready description of the capture batch (page counts, viewports, capture mode, any failures) so the generated images can be handed to an AI system for the visual review.

A typical workflow


ShotSweep
   ↓
screenshots
   +
manifest
   ↓
AI visual review

Run shotsweep capture ... --describe.
Attach the files listed in manifest.json, plus the printed description, to your model of choice.
Ask it to review layout, responsiveness, and consistency across the batch.