vlmkit

name

vlmkit

description

Automatic frontend quality router. Use automatically whenever the user asks to create, edit, debug, validate, test, compare, migrate, or repair a frontend UI, HTML/CSS, screenshot implementation, responsive or interactive behavior, Playwright/VRT, or a visual regression. The user does not need to mention vlmkit or choose a sub-skill. Classify the request, load the bundled workflow, run the smallest deterministic gates, fix failures, and rerun to green.

vlmkit — Skill Router and CLI Guide

Overview

This is the single automatic entry for vlmkit's focused agent workflows and the reference for its Visual Regression Testing (VRT) and accessibility verification CLI. The user installs this skill once and describes the outcome they want in ordinary language.

Automatically verifies that changes are visually and semantically (a11y) as intended, running a loop to detect and repair regressions.

Automatic routing contract

1. Do not ask the user to choose or name a specialized skill. Infer the workflow from their request, supplied artifacts, and files in scope.
2. Resolve every workflow path below relative to the directory containing this SKILL.md, not relative to the user's project. Read the selected workflow's SKILL.md completely before editing or running its gates.
3. Default to markup-assist for ordinary frontend work when there is no stronger screenshot, behavior, test-generation, comparison, migration, or benchmark signal.
4. Select one primary workflow. Chain a second only when the first workflow's output is genuinely the second workflow's input—for example, recreate a mock with mock-markup, then prove an explicit motion brief with dynamic-markup.
5. Ask only for a missing task artifact that materially blocks the work (such as the target screenshot), never for a skill-selection decision.
6. Execute the selected workflow; do not stop after recommending a command. Finish only when its stated deterministic done conditions are green, or report concrete evidence that blocks them.