Context: Stochastic Loot Engine

Problem Domain

This system is a stochastic loot engine for an idle or simulation game. The core loop generates unique items and skills with randomized modifier ranges, governed by rarity tiers, and evaluates character loadouts for emergent power interactions. The central design axiom is that overpowered item and skill combinations are not design failures but the primary attractor state of the game loop. The generative system must make such combinations discoverable, reproducible, and shareable while ensuring they remain statistically rare. Canonical game-design worldbuilding for the game this engine serves lives in docs/lore/README.md, the source of truth for game-design intent.

Cynefin Domain Classification

Component	Domain	Rationale
Skill Primitive	Clear	Formal syntax with unambiguous production rules and static verification properties.
Charge Resource System	Clear	Deterministic formula with bounded behavior and explicit state transitions.
Modifier Aggregation	Clear	Two-layer commutative-semiring structure with mandatory cross-layer ordering.
Item Template Generation	Complicated	Multi-parameter schema with covariance constraints; analyzable but requires expertise.
Rarity Distribution	Complicated	Beta-parameterized tier system with monotone nesting laws; tractable but non-obvious.
Condition Temporality	Complicated	Three-class taxonomy (spatial, decaying, accumulating) with evaluator registry pattern.
Trigger Graph Engine	Complex	Emergent graph structure from effect-to-trigger matching; cycle detection is algorithmic but outlier classification is emergent.
Stochastic Verification	Complex	Monte Carlo simulation with importance sampling; outcome distributions are not analytically derivable.
Outlier Detection	Complex	Phase-transition behavior in supercriticality; requires empirical probing to calibrate thresholds.
Algebraic Laws	Clear	Static properties amenable to property-based testing; violations indicate defects.

Scope Boundaries

In Scope

Core domain primitives: Skill, ItemTemplate, ChargePool, ConditionTemporality, Ascender
Generation layer: template generation from schema, instance rolling with Beta distributions, rarity tier assignment
Verification layer: static structural checks, compositional trigger graph analysis, stochastic simulation with Monte Carlo sampling
Algebraic law verification via property-based testing
Loadout classification schema (None, Candidate, VerifiedOutlier, UberOutlier) and outlier detection

Out of Scope

Game client, rendering, or UI presentation layer
Network multiplayer or server infrastructure
Save file persistence beyond seed-based reproducibility
Real-time combat simulation, positional physics, or movement systems
Non-loot progression systems (experience, levels, achievements)
Cross-session cooldown persistence design (decision recorded, not implemented)
Outlier presentation layer (what the player sees when a build is classified as outlier)