qa

Run the static gate (tsc --noEmit, npm test, vite build) and then a runtime gate: execute the app and perform the actual user action, smoke-test API endpoints against real data, and run SQL against real rows. Only mark done after observing correct behavior in the running application.

Check that all UI text is Hebrew including tooltips and error messages, dark mode renders correctly, fonts are readable at distance, RTL layout does not flip elements incorrectly, and destructive actions require confirmation. Also verify event logging, soft deletes, and no leftover console.log or TODO statements.

Static checks prove code compiles and pure logic is correct, but cannot catch runtime, visual, or wrong architectural assumptions like incorrect route prefixes, misrouted event handlers, or wrong DOM structure assumptions. Runtime verification against the live application is required to catch these.

When interactive verification is impossible due to login, map, or hardware constraints, the finding is marked as a critical blocker requiring user testing, not a minor side note. The done status is conditional on the user confirming they saw it working.

No. Without passing the runtime verification gate, the maximum recommendation is release with reservations, never ready for release. Compiling code and passing tests alone is necessary but not sufficient for approval.