k8s-acceptance-loop

What problem does it solve?

This Skill helps teams repeatedly verify that a user-visible browser UI behavior matches what should happen when running against the local ax-next-dev kind cluster, and then keep applying changes until the observable outcome passes.

Core Features & Use Cases

• Browser-observable acceptance loop: Drives the chat (or named UI surface) using Playwright MCP, then checks the DOM, console, and network responses for PASS/FAIL.
• Two-iteration modes for speed: Uses a fast hostPath dist mount loop for host-side TypeScript-only changes, or an image-rebuild loop for Dockerfile/chart/env/runner-side code changes.
• Structured triage across layers: Collects browser snapshots plus cluster-side signals (host logs, runner pods, init jobs) to classify the likely root cause and choose the right loop mode.
• Precondition setup for kind: Ensures the kind cluster, loaded images, runner namespace, and helm install are in the expected state before iterating.

Quick Start

Tell the Skill what browser-visible outcome should appear in the UI against ax-next-dev (for example, “after sending a chat message, the assistant reply contains the tool output X”), and the loop will drive the scenario in a real browser, capture evidence, apply the right rebuild or fast-loop update, and re-check until it passes.