Arcana Trigger: Hybrid Deck-Building Action Roguelike
Motivation
Arcana Trigger combines turn-based deck-building with real-time top-down shooting. A player can switch to action combat when a difficult hand limits their options, or use card strategy when dodging and aiming are challenging. Both modes share health and status effects so that switching changes how the player fights while preserving the same battle.
Co-designed and developed by a two-person team over approximately two weeks for the NHN Game × AI Hackathon 2026 preliminary submission, this project broadened my experience from game AI research to implementing a playable game’s rules, interfaces and progression. I was responsible for card game design, card combat, floor-level procedural generation and related integration.
Play the Web Demo · Source Code
My Contribution
- Core and card combat: deck, hand, discard, energy and turn resolution; shared battle state and card/action session transitions.
- Cross-mode rules: weapon-dependent attack cards, shared burn and slow effects, enemy intentions and boss card patterns.
- Floor-level procedural generation: implemented PCG that randomly generates the entire level on each floor, including seed-based room and corridor generation, connectivity validation, obstacles, exploration visibility and integration with the game scene.
- Progression and presentation: card rewards and removal, combat HUD, floor transitions, endless-run scoring and game-over results.
- Integration and verification: documented rules and interfaces, scene integration, and focused Unity EditMode/PlayMode regression checks.
My teammate primarily handled movement, aiming, shooting, enemy action behavior and bullet patterns. We coordinated around shared rules and game-scene integration.
Engineering Approach
The central challenge was preserving one battle state across two different time models: turn-based cards and real-time action. Explicit session lifecycles and state-transition rules kept health, effects and rewards consistent across mode changes. Seed-based dungeon generation and connectivity checks made layouts reproducible for debugging.
Development & Verification
The two-person team used Unity 6 and C# and deployed a WebGL build to GitHub Pages to provide a web demo.
During development, I used Codex Desktop for feature specifications, existing-code analysis, implementation planning, C# code and test authoring, development documentation and error analysis.
The developer reviewed AI-proposed code and changes and decided whether to incorporate them through diff review, playtesting, Unity Console checks, EditMode/PlayMode tests and visual inspection.
Development records include focused runtime, content, UI and scene checks. They do not establish that the final build and every browser path passed; some records retain font/content test failures and pending manual or Web validation.
Outcome & Lessons
The team submitted the prototype but did not advance beyond the preliminary round. The experience lies in taking a game concept through concrete mechanics, UI, procedural content and team integration. It also highlighted the need to keep feature implementation, automated checks, playtesting and final browser validation as distinct milestones.