Game Software Architect Specialist

Illustrative preview: this representative editorial result is not a recorded runtime session.

task_id: ARENA-ARCH-1
artifact_id: TIMING-SAVE-CONTRACT
version: 1
status: NEEDS_DECISION
producer / recipient: game architect / prototype owner and developer
objective / scope: Close simulation, one-shot input and checkpoint boundaries for the supplied arena rules.
inputs_and_sources: DASH-1, PAUSE-1, INPUT-1 and the checkpoint fields supplied in the request; no repository inspected.
confirmed_decisions: Offline 2D; cooldown uses simulation ticks; pause freezes simulation; menu animation uses a separate cosmetic clock.
recommendations_and_pending: Hold/repeat policy, post-load dash availability and input buffering while paused need their decision owner. Engine and tick rate remain unspecified.
criteria: DASH-1, PAUSE-1, INPUT-1, SAVE-1 proposed below
next_action: Prototype owner settles the pending input/load policies; developer can begin the resolved clock and one-shot consumption slice.

Architecture packet
Player path: capture press -> authoritative simulation accepts or rejects dash -> presentation reacts -> checkpoint serializes the defined durable state.

Proposed decisions

1. Simulation owns an integer tick and next_dash_tick. On a permitted dash at tick T, set next_dash_tick = T + 20. Permit again at that boundary, not before it. No seconds conversion or tick frequency is inferred.
2. Input capture queues one action token per physical press. The simulation consumes a token once, irrespective of catch-up tick count. Held input does not create extra tokens until its policy is settled. Alternative: consume a frame edge directly; simpler but risks reuse across catch-up ticks without an explicit consumed flag.
3. Pause stops authoritative tick advancement and rule effects. Menu animation follows a cosmetic clock. A single shared wall-clock timer is rejected because it conflicts with PAUSE-1.
4. The checkpoint codec validates position, health and stable pickup IDs before applying them to temporary state. In-progress dash is excluded. Whether load starts ready or starts a new cooldown is unresolved; do not silently choose.

Contract / owner / failure

• Clock: simulation scheduler; tick does not advance while paused; menu has no authority over gameplay.
• Dash gate: rule state; input is accepted only when the supplied eligibility rules and cooldown permit; repeated delivery must not issue a second dash.
• Input queue: input adapter; each press has one consumable identity; focus loss and paused buffering follow policies still pending.
• Save restore: checkpoint coordinator; validation precedes coherent state replacement; malformed data cannot partially overwrite live state.

Proof map, planned only

• DASH-1: dash at tick 100; tick 119 rejects, tick 120 permits. Rule-level trace.
• PAUSE-1: pause at tick 110; elapsed wall time cannot change tick/cooldown; separately inspect menu animation in the chosen engine.
• INPUT-1: one press followed by a render frame with several catch-up ticks emits at most one accepted dash. Input/scheduler integration probe.
• SAVE-1, proposed criterion: restore only the three durable fields atomically; post-load dash behavior waits for a decision. Rule codec plus scene integration and invalid-save probes.

First slice: implement the simulation clock and one-shot token consumption with an isolated rule fixture; defer unresolved hold/buffering and post-load dash behavior. A follow-up engine integration slice must prove input capture and animated menu behavior. No implementation, benchmark, engine run, controller test or playtest was performed; this is a proposed architecture, not a compatibility or fun claim.