Skip to main content

/gws-qa-run

Run QA test cases against the real app, in parallel, each in its own isolated environment. Type /gws-qa-run inside your AI agent and it spins up isolated gws forks, drives each test case, then publishes the run so you can read the pass/fail report and acceptance-criteria coverage in the GetWebStack dashboard.

This is the second half of the Branch QA Report loop. Generate and review the cases first with /gws-qa-generate.

When to use it​

  • You have a reviewed list of test cases (from /gws-qa-generate or written by hand) and want to run them
  • You want each case exercised against a real, production-parity stack with controlled seed data — not a shared URL
  • You want a shareable report before a branch merges

What it does​

  1. Isolates each case — runs cases in parallel across isolated gws forks so they can't interfere with each other. Cases that are provably read-only share a single fork to save resources.
  2. Drives the real app — each case runs against the full multi-service stack in its own namespace, with controlled seed data, so failures reproduce deterministically.
  3. Captures evidence — pass/fail per case, screenshots, screen recordings, and repro steps, plus the suspected failing code area for failures.
  4. Publishes the run — uploads the run (runs → cases → steps → artifacts) to the API so the full history lives in the dashboard, not just a local report.html.
  5. Surfaces coverage — shows how the results map back to the branch's acceptance criteria.

After the run​

Open the run in the GetWebStack dashboard to read the report and coverage, share it with reviewers, and decide whether the branch is safe to merge. The curated cases + seed + env config persist per app and compound over time.

Why isolation matters here​

Because GWS owns the system under test — it builds the environment from your repo — each case gets a fully isolated full-stack environment with deterministic seed data. That's what lets many cases run at once and lets a failure be traced to a suspected code area, rather than just "the page broke."

See also​

  • /gws-qa-generate — generate and seed the cases this skill runs
  • Branch QA Report — the end-to-end use case
  • gws qa — the CLI that publishes runs and syncs cases with the API
  • gws fork — the isolated worktrees each case runs in