What problem does it solve? Rambling bug reports like "it feels broken" either get filed as one unactionable mega-issue or discarded entirely. This Skill decomposes vague QA feedback into discrete, durable, independently-fixable issues worded in domain language, without over-interviewing the reporter or asserting unverified claims. ## Core Features & Use Cases - Conversational intake: Listens to plain-language bug descriptions, asks at most 2-3 clarifying questions, and files one thin issue at a time. - Mode-switchable filing: Defaults to a local ISSUES.md punch list for solo work, escalating to gh issue create with dependency ordering and AI-provenance notes only for shared GitHub repos. - Vibe decomposition: Breaks "everything feels broken" anchors into separate behavioral issues with What happened / Expected / Steps to reproduce / Context structure, leaving unverified doubts as open questions for triage. - Use Case: A beta tester texts that a salon booking app "feels broken and cheap" on mobile. The Skill files three thin issues (overlapping slot-picker chips, a hanging Confirm spinner, an unbranded confirmation email) and flags email delivery as an open question rather than an asserted bug. ## Quick Start Run a QA session on my app: I'll describe what's wrong in plain language, and you file each problem as a thin issue in ISSUES.md.