Структурирование задач для ИИ-агента

★ 8.2 · ai-agents

dbs-good-question is a Claude Code skill that rewrites vague problems, observations, and hunches into structured problem specifications an AI agent can reason about, critique, and verify. It operates across four modes: reformulating fuzzy questions into concrete specs with a defined object, goal, conflict, constraints, and feedback loop; generating 2–3 candidate explanations evaluated against a "hard to vary" criterion; assessing agent solvability across six dimensions (boundary clarity, expressible variables, feedback availability, testability, executable actions, pattern stability); and rating problem clarity as low, medium, or high before outputting either a draft problem spec, a minimal set of follow-up questions, or a four-tier automation verdict. Typical use cases include diagnosing content conversion drop-offs, sales funnel breakdowns, and business process automation feasibility — anywhere a question needs to be sharpened before an agent can produce reliable analysis.