A local visual QA gate for AI-built software. FrameCritic checks real browser interfaces and rendered product videos before an AI coding agent is allowed to call the work finished. It captures visual evidence, detects layout and composition failures, and exports an agent-ready fix packet so builders can move from “the agent says it passed” to proof they can actually inspect.
Now im testing it
Deep-tested FrameCritic v0.2 tonight and continued the Windows compatibility audit. Fixed PowerShell test quoting so the full test suite runs correctly, cleaned up v0.2 version/report inconsistencies, improved responsive sweep errors, and tested paths containing spaces plus Windows artifact handling. FrameCritic is now at 179 passing tests. Next I’m continuing dogfooding, packaging, security, and report-quality checks.
Just polishing it
