What problem does it solve? Engineering tasks often stall or go wrong because critical facts are missing, yet asking broad or redundant questions wastes time. This Skill identifies only the factual gaps that materially affect correctness, scope, testing, rollout, or runtime behavior, asks the minimum set of questions, and records confirmed facts in the task repository so context survives across sessions. ## Core Features & Use Cases - Minimal question generation: Filters out questions answerable from code, tests, docs, or existing task artifacts, and groups remaining questions by category (business rule, contract, schema, runtime, tests, rollout). - Task state persistence: Updates TASK_STATE.md with open blockers, confirmed facts, and next steps, and writes to DECISIONS.md only when evidence-backed resolutions exist. - Unattended-run safety: In background or fleet-dispatched runs, records open questions as [NEEDS CLARIFICATION] markers instead of inventing answers. - Use Case: After an impact analysis reveals uncertainty about an API payload contract, run this Skill to produce three targeted questions, update TASK_STATE.md with the blockers, and route to the correct next command once answers arrive. ## Quick Start Ask the agent to run targeted-questions on the active task to identify the minimum factual questions needed before planning can proceed safely.
