Drill: Command palette: βK palette with keyboard nav (ββ Enter Esc), grouped results, fuzzy match, footer hints. Must work without mouse.
Interpretation: One cue palette for a stage-management desk (Mothlight Theatre, tech rehearsal R3), shown at six moments of one keyboard interaction β idle standby set β fuzzy 'moth' β cross-group navigation β no-results recovery β '@' files mode β GO receipt. Every frame is the deterministic output of a real contiguity-bounded fuzzy/selection engine; a hidden live mount receives synthetic keydown events so 14 on-load audits prove ββ/Enter/Esc/mode behavior without a mouse. Warm-paper prompt-book register, Inter + JetBrains Mono, single GO-green accent reserved for selection/match/live/GO semantics.
Tech rehearsal R3 Β· Wed 28 Oct Β· 19:42 Β· book held by DSM R. Okafor Β· decks: LX D. Whitcombe Β· SND P. RΓ©ti Β· FLY M. Ashworth
Each panel below is the actual output of the palette engine after the key sequence shown β not a mock-up. Fuzzy match is a case-insensitive subsequence over label then description; ββ cross group boundaries; β fires; esc clears the query, then closes. Mode prefixes scope the search: @ files Β· # cues Β· > commands Β· / channels.
Correct when β the three standby cues lead the list, exactly one row is selected, zero match marks, footer hints present.
Correct when β every visible row matches "moth", matched glyphs are marked in place, group counts equal their rows, rank-1 selected.
Correct when β selection sits on the first channel row (rail + fill + β chip on it alone); LX 47 shows its STANDBY tag again, not β.
Correct when β zero rows, the query is echoed, three recovery chips β₯44px, and the naming rules (department prefixes) are stated.
Correct when β a "@ files" chip sits inside the input, documents only are listed, counts are right, first document selected.
Correct when β the receipt names what fired and when, the next standby set, deck acknowledgements, and logs itself to the rehearsal report.
Fuzzy match: case-insensitive subsequence, label before description; label matches rank above description matches inside a group, and description matches carry a contiguity bound β the matched span may not exceed 3Γ the query length, so scattered hits across a sentence are rejected (βtimestamp the report at this cueβ is not βmothβ). Selection: ββ moves one row across group boundaries; β fires the selected row and prints a receipt; esc clears the query, a second esc closes; mode prefixes @ # > / scope the result set before matching. The engine is deterministic β all six frames above are its output, replayed from the recorded key sequences.
Verification: audit checks pending run on load in both viewports β they assert the frame row-sets against the engine, the single-selection invariant, the tagββ chip swap across a group boundary, the no-results recovery contract, mode scoping, the receipt's successor chain, and keyboard behaviour on a live hidden mount (synthetic ββ / β / esc events β no mouse).