Image models kept getting my App Store screenshots rejected, so I made my coding agent render them to exact spec

Short summary

The author built HyperShots, an open-source skill for Claude Code and Codex that renders App Store screenshots as deterministic HTML/CSS instead of relying on AI image models. Image models fail Apple's strict geometry and text requirements, so the tool uses Headless Chrome for pixel-exact rendering, a fail-closed validator for Apple's specs, and an optional AI style pass via fal.ai. Two production apps have shipped validated screenshot sets through the pipeline.

• •HyperShots renders App Store screenshots as deterministic HTML/CSS via Headless Chrome, solving geometry and text issues AI image models can't
• •Includes a fail-closed validator that checks dimensions, alpha channels, ICC profiles, panel count, and file size against Apple's actual rejection criteria
• •Optional generative layer adds AI sticker art and style passes while preserving pixel-true typography via re-compositing

Generated with AI, which can make mistakes.