Strategy interface review — 2026-09-09

The owner requested a broad interface pass toward a normal strategy game, with clear play decisions, unit movement, and enough detail for rules analysis. Work stays on feature/local-playtest-prototype for local testing. There was no merge, push or deployment.

Research and method

The interface research records primary sources and how they apply here. References include the Civilization VI manual, Into the Breach's public design description, and Xbox accessibility guidance. These references informed presentation. The answered owner interview remains the rules contract.

Three agents handled independent research, melee presentation, movement presentation, descent diagnosis, and review. The main agent integrated the battlefield and analyst tools. The agent that wrote the descent fixes did not approve its own engine changes; the movement agent reviewed those separately.

The first review scored the old interface 58/100 against this new quality bar. The next pass scored 80–82. Findings included an incorrect melee decision banner, hidden actionable pools, excessive empty board space, unclear results, and keyboard access behind Game tools. Implementation continued until both final reviews cleared the 90-point target without blocking findings.

Final review scores

The integration reviewer scored the interface 91/100:

The independent research/usability reviewer scored it 92/100 using its recorded orientation, actions, tactics, feedback, melee and analyst-access rubric. Both reviewers found no blocking issue in the flows they inspected and approved owner local testing. These are subjective interface assessments, not a user-study result or a balance verdict.

Changes reviewed

• Focused battlefield with stable dimensions, visible score, true decision owner and compact unit strip.
• Separate Orders, Inspect and Analysis panels. Inspection never changes the active mech.
• Move, Attack and Support commands with costs, visible unavailable reasons, rule explanations and explicit confirmation.
• Current-player melee pool first, grouped sources, target-part defenses, responsive parry cost, optional full dice ledger and persistent confirmation.
• Accepted-path movement animation, Skip, safe reduced-motion behavior, engagement framing and optional callouts. Historical events do not replay as live movement.
• Immediate hit, actual durability-loss and Heat feedback. The analyst view explains each roll through penetration, damage, critical effects and final state.
• Native modal Game tools with keyboard isolation, synchronized preferences and retained live settings when storage fails.
• Safe descent routing, accessible deliberate-fall planning, accurate destroyed-Jump reach and a stepped rooftop trial.

Validation

• pnpm test: 377 tests passed across 48 files.
• pnpm typecheck: passed.
• pnpm test:browser: 26 checks passed across desktop and tablet landscape.
• Browser coverage includes actual intermediate movement, Skip without command/dice changes, reload and reduced motion, safe 2 → 1 → 0 descent, melee actor and footer visibility, stable board bounds, modal keyboard isolation, shot explanations, saves, reactions, private deployment and sideboarding.
• Existing worked example J, seeded simulation and controller parity remain covered. No ranged RNG order changed.
• Packet data, generated rules and the production build are regenerated after the documentation pass.

Remaining scope

Small secondary text and repeated unavailable-action reasons can still receive polish. User testing should assess pacing, discoverability and the tactical value of the information shown. Melee balance, final art, remote play, phone portrait optimization, final cone geometry and the public call-sign convention remain outside this interface approval. The last two are explicitly provisional owner follow-ups.

Owner follow-up: turn start and Orders

The owner’s local testing found the player-start control difficult to locate and the order categories confusing. This follow-up supersedes the earlier Move / Attack / Support organization. Begin Player A/B’s turn now sits in the persistent status bar, with Respond as Player A/B in the same location for melee and reactions. Taking control returns to the correct decision panel without consuming an action. Ready-mech cards replace the duplicated round banner, and End activation stays at the bottom of the side panel.

Orders groups each command once under Movement, Ranged, Melee, Defense or Systems. All orders and category filters remain available. The selected order is named in the HUD even during inspection and has explicit return/cancel controls. Weapon controls are available before arming a shot. Review caught an invalid weapon selection after arm destruction; selection now falls back to an available slot and keeps the preview and actual accepted shot consistent.

The independent reviewer scored this follow-up 93/100, with no remaining blockers in desktop/tablet turn, melee, reaction and weapon-recovery flows. The breakdown was play flow 24/25, hierarchy 19/20, transparency 19/20, feedback 13/15, accessibility 9/10 and state safety 9/10. The remaining polish finding is lengthy unavailable reasons in All orders; category filters limit that density. This remains a subjective review for local testing.

Validation: 377 unit tests, 32 desktop/tablet browser checks, and typechecking passed. New browser checks measure persistent handoff and board bounds across every panel, verify that readiness spends no action, exercise grouping/selection/cancellation, and fire the sole surviving gun after importing a disabled selected arm. Existing melee, movement, reaction, private-deployment and replay checks remain green. No rules or engine mechanics changed in this follow-up.