What problem does it solve? Quantitative traders running daily batch inference on thousands of stocks need a repeatable methodology to turn raw fusion scores into actionable decisions: whether to enter the market, which stocks to buy, how much position to take, and which names to short. This Skill encodes that decision framework against the QuantMind API. ## Core Features & Use Cases - Market State Judgment: Three-layer filtering using industry signal strength (ind_avg_top1), index MA20 trend, and strong-industry counts to decide entry, position sizing, or staying flat. - Stock Selection & Short Reference: Score-band based picking (golden zone, chase-risk zones), 3-day score trend patterns, and negative-score short candidates filtered by market cap tier and board. - Score Calibration Backtest: Submits async score-calibration tasks per model to map absolute thresholds to each model's actual score distribution, since thresholds are not portable across models. - Use Case: After a nightly batch inference completes, ask the agent to analyze the latest batch: it pulls member runs, checks market state, lists 3-5 buy candidates in the calibrated score band, flags micro-cap deep-negative names as short references, and summarizes sector rotation over the past 30 trading days. ## Quick Start Ask the agent to analyze the latest completed batch inference results and output today's market state, position sizing, stock picks, and short-selling candidates.
