So far, our soldier's two main knowledge calculations are within enemy selection and flee position selection. With a proper KnowledgeSource instance implementation, we can refactor our original evaluator's internals to a standalone function that returns a result and confidence value in the format expected by our KnowledgeSource instance implementation.
Converting the logic from the SoldierEvaluators_HasEnemy function only requires modifications that return both a confidence value and the final result. Since only one source will determine enemy selection, we will return a confidence of 1, as well as the selected enemy.
SoldierKnowledge.lua:
function SoldierKnowledge_ChooseBestEnemy(userData)
local sandbox = userData.agent:GetSandbox();
local position = Agent.GetPosition(userData.agent);
local agents = Sandbox.GetAgents(userData.agent:GetSandbox());
local closestEnemy;
local distanceToEnemy;
for index=1, #agents do
...