triage-visual-changes

Triage UI Verify visual regression builds over MCP and accept baselines in bulk.

21|1|Updated Jul 22, 2026
One-click install
npx skills add https://github.com/uiverify/uiverify --skill triage-visual-changes-uiverify
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: triage-visual-changes
Source: https://github.com/uiverify/uiverify/tree/main/packages/skills/skills/triage-visual-changes
Command: npx skills add https://github.com/uiverify/uiverify --skill triage-visual-changes-uiverify

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When a UI Verify build reports visual changes on a pull request, reviewing dozens of changed stories in the dashboard is slow and error-prone. This Skill lets a coding agent bucket changed stories into real regressions, intended changes, cosmetic reflow, and rendering noise directly from the terminal via the UI Verify MCP server. ## Core Features & Use Cases - Build triage over MCP: Page through a build's changed stories with get_build and list_build_stories, read the AI judge's verdicts, and render baseline/candidate/diff images inline to adjudicate ambiguous changes. - PR-ready summaries: Produce a one-line-per-regression comment with before/after images and presigned diff URLs so reviewers never leave the pull request. - Bulk baseline acceptance: Accept every changed story in one accept_build call, or loop review_diff per story for targeted accept/ignore/deny decisions. - Use Case: A PR flags 67 changed stories. The agent pages through them, compares before/after pixels for the flagged ones, reports 2 real regressions and 11 intended changes, and posts the summary with cropped before/after images to the PR. ## Quick Start Connect the UI Verify MCP server with your project API key, then ask the agent to triage the latest UI Verify build for this pull request and summarize the regressions.

Frequently Asked Questions about triage-visual-changes

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I triage visual regression changes from a coding agent?▼

Connect the UI Verify MCP server, then call get_build for the PR to see changed stories with AI judge verdicts. Page remaining changes with list_build_stories, render baseline and candidate images for ambiguous stories, and bucket results into regressions, intended changes, reflow, and noise.

How do I accept all new visual baselines at once?▼

Call accept_build with the build's commitSha, prNumber, or buildId to accept every changed story in one call. Only do this after confirming no real regression hides in the list, since it makes the candidate the new baseline for all changed stories.

How do I connect the UI Verify MCP server to Claude Code or Cursor?▼

The server is a remote Streamable-HTTP MCP at https://uiverify.ai/api/mcp authenticated with a Bearer uv_proj API key. In Claude Code run claude mcp add with the header; in Cursor add the URL and Authorization header to mcp.json.

Can I see before and after images for a story that did not change?▼

Yes. Pass storyId to get_diff to get the current baseline URL even on a passed build, or use render_diff_image with storyId and which baseline for inline pixels. Enumerate unchanged stories with list_build_stories status unchanged.

Why should I render both baseline and candidate images instead of just the diff mask?▼

The diff mask only shows where pixels moved, not whether the result is correct. Issues like illegible contrast, clipping, or overlap only appear when comparing before and after images side by side, so judging the candidate alone can pass real regressions.

When should I not bulk-accept a UI Verify build?▼

Avoid bulk accept when any story is bucketed as a real regression or when a global CSS or theme change swept into stories the PR never intended to touch. Use review_diff per diffResultId for targeted accept, ignore, or deny decisions instead.