Sync one manual test with the code, run it in the cloud, and get pass or fail with video
The sync-and-run skill takes one manual test and does the whole loop: check the test against the current code, update it when a step is out of date, then run it as a UI test with a walkthrough video.
Use it as the base for the nightly QA dashboard: one run per test, all tests in parallel, one red / yellow / green row per test with its video.
Type /sync-and-run-test in chat and name one file under tests/manual/.
When the run is for an issue, also name the Jira key.
The local launcher lives at .cursor/skills/sync-and-run-test/. One command
runs the whole sequence and prints the report when it ends. The run can take
hours; the laptop must stay on until it exits.
The cloud path needs no laptop. Start a run against the dev deployment with the team's shared secret, and read the result back by run ID.
The agent compares every code-facing claim in the test with the code. When a
claim is stale, it changes only the test file and pushes it to dev, retrying
once when dev moved. When the push still fails, it opens a pull request and
the UI test does not run.
The agent runs the test in the browser with fresh accounts and inboxes, then records the walkthrough video and uploads it. Only the video survives; the HTML report is deleted with the box.
Each run reports:
The local launcher has a self-test with stub phases: pass, sync fail, and pull request. It fails when a run uses more than 12 steps.
The cloud workflow and trigger live in @zooly/ui-test-runner-srv, with thin
routes in zooly-app. The boot scripts and agent prompts are ported from the
local sync and UI skills.
The dashboard exists: the test console starts runs and reads their reports from the app database. One run per test file, all tests in a suite in parallel, one red / yellow / green row per test with its video.