What problem does it solve? News cards generated from ingested articles often contain thin prose, false facts, or missing background, and there is no repeatable way to measure and fix those gaps. This Skill provides a structured iteration loop that runs the live news cycle, displays every card verbatim, diagnoses the worst gap against a fixed target card, and coordinates fixes until quality converges. ## Core Features & Use Cases - Repeatable iteration loop: Capture fixtures, run the live cycle (poll, synthesise, enrich, build cards), show every card whole, fix the worst gap at its cause, merge, and repeat. - Fixed measuring stick: Every run is compared against one hand-written target card rather than a moving score, so progress is judged by closeness to a concrete example. - Coordinated fixes: Small fixes happen on the main thread while disjoint-file tracks are dispatched to background sub-agents, with regression floors riding the test suite. - Use Case: An operator says "run iterations 14 to 18 of the news loop" and the session runs the cycle, surfaces thin cards and false facts, fixes the extraction or enrichment defect responsible, and stops when the budget or a no-change condition is reached. ## Quick Start Tell the session to follow the news-feed-quality skill and run the next iteration of the news card loop.
