refactor(dataclass): remove pyright ignores by using parameterized generic aliases as field factories

[Copilot]

field(default_factory=list) causes Pyright strict mode to infer list[Unknown] regardless of the field's type annotation, requiring # pyright: ignore[reportUnknownVariableType] suppressions. Using the parameterized form list[T] / dict[K, V] as the factory gives Pyright the exact return type it needs.

Changes

• 8 fields across 5 files — replaced bare list/dict factories with their parameterized equivalents and removed all associated ignore comments:

File	Field	Factory
kagent/state.py	tools: list[Tool]	list[Tool]
kagent/trace/entry.py	data: dict[str, Any]	dict[str, Any]
kcastle/providers/config.py	options: dict[str, object] ×2	dict[str, object]
kcastle/providers/config.py	models: list[ModelConfig]	list[ModelConfig]
kcastle/config.py	options: dict[str, object]	dict[str, object]
kcastle/config.py	providers: dict[str, ProviderEntry]	dict[str, ProviderEntry]
kcastle/skills/skill.py	tags: list[str]	list[str]

# Before — Pyright infers list[Unknown], ignore needed
tools: list[Tool] = field(default_factory=list)  # pyright: ignore[reportUnknownVariableType]

# After — Pyright infers list[Tool] exactly, no ignore needed
tools: list[Tool] = field(default_factory=list[Tool])

The remaining # pyright: ignore[reportUnknownVariableType] comments in the codebase (on dict-iteration over untyped YAML data) are a separate, unrelated issue and are not touched here.

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.