Compare baseline and current screenshots to write per-route vlm_verdict and vlm_summary into index.json.
npx skills add https://github.com/paulingham/.claude --skill vlm-critic
Please help me install this Agent Skill. Skill: vlm-critic Source: https://github.com/paulingham/.claude/tree/main/skills/vlm-critic Command: npx skills add https://github.com/paulingham/.claude --skill vlm-critic
Final Gate teammate for the vlm-critic agent orchestrates visual-regression evaluation by reading index.json routes and plan.md, comparing baseline and current PNG pairs, and writing per-route vlm_verdict and vlm_summary back into index.json, then returning an aggregate verdict of VISUAL_DIFF_PASS or VISUAL_DIFF_FAIL to gate the product-release.
Spawn the vlm-critic during the Final Gate to produce per-route visual-regression verdicts by reading index.json and plan.md, and write vlm_verdict and vlm_summary back to index.json.
High-intent search queries and answers about installing and using this skill.
To detect semantic visual regressions, compare baseline and current PNG screenshots across defined routes and generate a PASS or FAIL verdict for each one. This process evaluates visible deltas and writes concise summary notes directly into your index.json file to gate the release.
A Final Gate visual regression check is a strict automated review step that compares captured design-qc screenshots against baselines. It enforces deterministic outputs and read-guards multimodal access, returning an aggregate VISUAL_DIFF_PASS or VISUAL_DIFF_FAIL outcome to gate product-release readiness.
You write per-route verdicts by reading index.json routes and plan.md, evaluating the screenshot pairs for each route, and updating index.json with a vlm_verdict of PASS or FAIL alongside a short vlm_summary to document the semantic diff findings.
Yes, you can restrict multimodal file reads by enforcing a vlm-critic-read-guard. This strict read-guard rule limits multimodal access to only the necessary baseline and current PNG pairs, ensuring deterministic evaluation while reading and writing exclusively to index.json.
Yes, this visual regression gating approach works without external dependencies. It operates autonomously by reading design-qc inputs like index.json and plan.md, enforcing deterministic outputs, and relying entirely on internal read-guards to restrict multimodal reads.
The limitation of using index.json for semantic diff gating is that the process reads and writes strictly to index.json, meaning all baseline and current PNG pairs must be mapped across index.json.routes beforehand for the deterministic verdicts to execute successfully.
caidishCapture the current KLayout layout and render it as a PNG.
wpgauravConvert product page URLs into ACF product-box block markup.
hffmnnjCompare screenshots pixel-by-pixel to detect unintended UI changes.
nickhou1983Automate visual regression review for Playwright tests using GPT-4o vision.
