What problem does it solve? Numerical render gates can pass while an image still looks wrong, and manually eyeballing every render output is slow and inconsistent. This Skill uses multimodal image reading to quickly flag visual problems in path tracer outputs. ## Core Features & Use Cases - Qualitative render inspection: Reads a render output and reports on exposure, contrast, colour balance, and overall impression. - Artifact detection: Flags fireflies, banding, NaN pixels (magenta/black spikes), aliasing, and moire in rendered images. - Reference comparison: When given a reference image, describes what changed, whether it improved, and any regressions. - Use Case: After a spectral pipeline change, run /visual-check on benchmarks/showcase/output/cornell_spectral_512spp.png to confirm dispersion and material fidelity before trusting the numerical gates. ## Quick Start Ask the assistant to run /visual-check on a render output path, optionally with a reference image path, to get a short inspection note.