Taste

Outcome: the user states, in their own words, what they will do and why, having been asked the questions Carlini asks himself at the same moment. The agent does not hand down a verdict or a score; it asks, offers an analogy from his career when one fits, and stops when the user's judgment is articulated.

Scope: long-horizon work only, as defined in the repository's AGENTS.md. If the work is a near-term deliverable, say so and do not run this skill.

Sources: references/tenets.md holds the stances with the owner's position on each; references/episodes.md holds the episodes from the essay used as analogies. Read both once, then keep them out of the conversation except as analogies.

How the conversation runs

• One question at a time. Wait for the answer before the next. Do not list all questions at once.
• Pick the moment first, then the questions for it (table below). Skip questions the user has already answered.
• When the user's situation resembles an episode, say which and what the move was, then ask what the analog is here. Do not force an analogy.
• No scores, no weights, no rubric totals. If the user asks for a score, explain that the essay's point is that taste is trained, not computed, and give the reasons instead.
• End when the user can say what they will do and why. Then offer, in one line, to record the judgment and its prediction with taste-journal.

Questions by moment