island-selectlisted

# island-select Goal: - Select the next island strategy and parent hypothesis set for one evolution round. Inputs: - `islands/ISLANDS.json` - current viable `hypotheses/<id>/HYPOTHESIS.json` frontier - `iteration_count` - numeric island-selection settings from `state/RESOLVED_RUN_CONFIG.json` Outputs: - deterministic `SelectionResult` - selection strategy - selected parent hypotheses - when consumed by routing, updated `state/STRATEGY_PLAN.json` signals Context Loading: - Open `skills/shared-references/schema-index.md`. - Read `packages/agent_contracts/state.py` and confirm the exact `IslandStateContract` shape used inside `islands/ISLANDS.json`. - Read `packages/agent_contracts/hypothesis.py` for the exact `HypothesisContract` shape. - Read `packages/agent_contracts/resolved_config.py` for the island numeric settings. - When the result is persisted into routing state, also read `packages/agent_contracts/strategy_plan.py`. Execution Contract: - This skill is deterministic and must not call an LLM. - Use `from tools import select_island_hypotheses` as the stable invocation surface. - The exported helper is implemented in `packages/agent_mechanics/island_select.py`. - The helper signature is `select_island_hypotheses(islands, hypotheses, iteration_count, ucb_exploration_constant=..., softmax_temperature=..., stagnation_epsilon=...) -> SelectionResult`. - The helper is the canonical implementation of UCB island selection, stagnation-triggered multi-island fallback, and