Step ① — Clarify what you are really doing

Why this exists: it is easy to do the right thing in the wrong direction. Before exploring how, fix what counts as the problem solved — otherwise every later step is efficient motion toward the wrong place.

Done = all three pieces are pinned

1. The real goal — what, and why. Chase it to the real goal: a means is not a goal. "Build an oracle" is a means; "the extractions are accurate" is the goal. The current task is usually a means to a higher goal — ask one level up: this artifact serves what?

2. Constraints + key assumptions — each rated importance × certainty. Include dependency constraints, and run the circular-dependency check: your measure must not depend on the thing it measures. (Judging extraction quality by the recall of that same extraction is circular — it will tell you what you want to hear.)