Bad Company → Good Stock

bad-company-good-stock automates a Korean value-investing playbook: find tiny, ignored, low-PBR companies whose owners are sitting on retained earnings and cash, where some catalyst (inheritance tax, governance dispute, embezzlement settlement) will eventually force them to return capital to shareholders. The skill scores every candidate across six conditions — micro/standalone/low-PBR, retained earnings ratio, cash ratio, controlling-shareholder ownership, governance/inheritance events, and obscurity. Each condition is a 0-5 rubric with explicit thresholds, so two analysts running the same data converge on the same score. It runs in three modes: batch screening (DART OpenAPI + Naver Finance + Python scorers), single-stock deep-dive (a 30-point report with star rating, risks, and catalysts), and web-search fallback when no API key is available. Candidates discovered in previous runs accumulate in `references/known_candidates.md` so the next run can reuse them rather than rediscover.