Automated typing tests

Prove Mac/Linux-style editing on the tablet with scripted tests — not by typing in the Lobby by hand.

Read: this file, milestone-runs.md, lessons.md, decisions.md (Typing-test strategy is failing, then §13), terms.md. Check names: scenario-catalog.md.

Test strategy status (do not remove yet)

Same rule as decisions.md Typing-test strategy is failing: we do not have a working strategy yet — only an attempt that keeps going green while basic editing bugs still reach a person. Keep this banner until that decisions section is lifted with solid proof the misses have stopped recurring. Method theory (update when we learn a new failure mode): methodology-shortcomings.md. Claim kill-tests: basic-claims.md.

Current score

All 125 checks passed at 21-43-24 (fork commit 19792fc). Basic set 57 of 57 (21-32-07): wrap Home/End, End-then-Down kill-test, goal-column Up round-trip. Do not run edit-session at the same time as these typing tests. Treat this score under the status section above.

Next

Owner Physical Home check (user-should-test.md). Keep all typing checks passing on any future edit change: one push, one deploy, edit-session, full --fast, update the pass/fail log. When a human finds a basic bug that was green, fix the test’s ability to catch that failure and update basic-claims.md before calling the strategy improved. Priority raises are listed at the bottom of that inventory.

Do not

Infer the moving end of a selection from the caret after select. Treat keyboard arrows as page-scroll. Deploy once per failing case. Auto-release Escape in key-feed helpers. Step wrapped lines by a tall caret rectangle alone.

Done means

Basic set 57 of 57, full set 125 of 125, edit-session pass, clean journal after deploy.