chenmoranzhiandCursor
5c097caee1
核心不得依赖引擎界面，随机数走统一入口，避免规则写进表现层。 Co-authored-by: Cursor <cursoragent@cursor.com>
16 lines
1.0 KiB
Markdown
16 lines
1.0 KiB
Markdown
# Agent rules
|
|
|
|
Read `docs/FOUNDATION.md` and `docs/ARCHITECTURE.md` before large refactors.
|
|
|
|
- `Game.Core` must not depend on Godot.
|
|
- All randomness must go through `GameRng`. Simulation and presentation share one rules engine.
|
|
- UI must not implement game rules. Presentation renders `GameState` and plays `GameEvent`s.
|
|
- Adding or changing core rules requires adding or updating tests.
|
|
- Ordinary cards should use standard effects. Do not invent a general-purpose scripting language.
|
|
- Changing the content schema requires updating the validator and `docs/CONTENT_SCHEMA.md`.
|
|
- Changing event resolution order requires checking related tests.
|
|
- Do not create a global rule for a single card.
|
|
- Prefer the smallest working implementation. Do not optimize until there is a measured bottleneck.
|
|
- Freeze the 11 core actions in `docs/RULES.md`. New “archetypes” are combinations, conditions, or keywords — not new primitive events.
|
|
- Card packs are **in-run only**. Unlock state must reset with the run. Do not add cross-run / meta unlocks.
|