Unit commands

The player selects one or more units and gives explicit orders (move here, attack this, use that, stay, guard) that take precedence over the units' autonomous decision making. State owned: per-unit order queue and override flag. Update rule: when an order is present, the unit's action selection is bypassed or constrained to executing it; on completion or cancellation, autonomy resumes. Orders can be queued and may be refused if impossible. In simulation games it often sits beside autonomy: RimWorld's drafting switches a colonist from autonomous work to direct control for combat; The Sims lets the player queue actions for a Sim who otherwise acts on free will; Dwarf Fortress squads take orders while civilians do not. Strategy games make it the default control mode.

When autonomy fails or stakes are high, the player takes the wheel. Drafting colonists into a defensive line or forcing a Sim to bed is precise and reassuring, but heavy use turns agents into puppets and the player into a micromanager.

• scope: any unit, only military units, or only drafted units
• how long overrides last and how autonomy resumes
• order queue length and shift-queue support
• whether units can refuse or delay orders based on morale or needs
• group orders and formations
• what autonomous behavior continues under orders, such as eating or fleeing

• Orders are needed constantly because autonomy is weak, turning play into micromanagement.
• Drafted or ordered agents ignore needs and starve or collapse, producing deaths players blame on the interface.
• Unclear return-to-autonomy rules leave units idle after an order completes.

• A dual mode emerges: autonomy in peace and direct command in crisis, giving play a rhythm of calm and intensity.
• The cost of micromanagement pushes players to improve autonomy-affecting systems such as priorities and schedules.
• Override on one agent shifts load to others, revealing hidden dependencies in job systems.

• Can units interpret high-level natural-language orders (hold the bridge until the medic arrives) and decompose them, and how should they report failure?
• When agents have language-model reasoning, how should they negotiate or refuse orders without frustrating the player?