Blueprint kit: phone app tests

5 scenarios for the unreleased phone app update (kit 1.1 draft), sent through the kit's real prompts. Each one is graded by a separate AI call against a pass rule written before testing. The security review is scored against a hidden answer key for a flawed Expo app, and a clean web app checks that the new phone checks don't cause false alarms. A scenario only passes if every run passes.

This uses your own Claude usage. 5 tests at 3 runs each is 30 calls and can take 10 to 20 minutes. Start with 1 run to check everything works.

Runs per test Tests

–scenarios passed every run

–individual runs passed

–hidden flaws found (all runs)

–false alarms on the clean app

Results to publish

Everything from this session: every raw answer, every grade and the kit version. Copy it into the tests repository's results folder, or paste it back into the chat.