Decide
Find the uncertainty that can change the outcome, authority, scope, reversibility, or credible verification. Resolve it with useful evidence: repository inspection, a reference, a focused experiment, a user decision, or independent analysis. There is no prescribed order among these methods.
Compare alternatives only when they differ on a meaningful axis. Explain the fact under which each viable option wins. When one option meets the requirements with less complexity and no hidden caller cost, choose it.
Preserve settled decisions unless new evidence invalidates their basis. Mark assumptions as assumptions. Ask for a material preference or missing fact the environment cannot provide; do not turn readable facts into an interview.
Start authorized work when remaining uncertainty is reversible or best resolved through implementation. Independent contexts are useful for premise challenges and expensive decisions; pass facts and decisions separately from hypotheses.
Keep continuity in chat or existing project docs. When a durable handoff helps, record the goal, settled decisions, boundary, acceptance, and unresolved question in the project's natural format. Task notes offer an example, not a required document grammar.