AI agent skill · by @MRCalderon3D
Input Abstraction
Design input around player intent and supported device families instead of hardcoding device-specific behavior everywhere.
Archived 2026-10-23 · Source updated 2026-08-20 · License: MIT. Check the original repository for the latest version.
SKILL.md instructions
# Input Abstraction ## Purpose Design input around player intent and supported device families instead of hardcoding device-specific behavior everywhere. ## Use When - multiple input devices are supported - rebinds or accessibility matter - UI and gameplay input are becoming tangled ## Inputs - action list - supported devices - UI flow - accessibility requirements ## Process 1. define actions by intent 2. separate raw device input from gameplay/UI action handling 3. document action states and context switching 4. support rebinding and persistence where needed 5. test edge cases around focus, pause, and overlays ## Outputs - action map - context model - device support notes - rebinding requirements ## Quality Bar - gameplay consumes named actions, never raw device inputs - rebinding, device switching, and multiple control schemes work without touching gameplay code - input contexts (gameplay, menu, cutscene) are explicit, with defined push/pop behavior - dead zones, buffering, and repeat behavior are configured per action, not hardcoded ## Common Failure Modes - raw key or button checks scattered through gameplay code - input consumed by multiple contexts at once (menu and gameplay both react) - device-specific assumptions that break on controller or touch - buffering and coyote-time tuning embedded in mechanics instead of the input layer ## Related Agents - ui-programmer - accessibility-reviewer - gameplay-programmer ## Related Commands - ui-flow-review - verify - tech-design ## Notes - Keep this skill aligned with the relevant rules layer and current project documentation. - If engine-specific constraints materially change the workflow, hand off to the matching engine skill or engine-specific reviewer.