backtest-runnerRun a historical simulation of a trading strategy on prediction-market data and report ROI, hit rate, Brier score, and per-cohort performance. Use when the user asks "run a backtest", "test this strategy on history", "see if the cap raise pencils out", "validate against historical prices", "would th
calibration-auditAudit how well-calibrated a probabilistic forecast is — does "70% confident" actually come true 70% of the time? Use when the user asks "is my model calibrated", "reliability diagram", "audit my predictions", "Brier score breakdown", "are my probabilities trustworthy", "should I trust this model's c
drawdown-monitorMonitor a strategy's running drawdown and halt or de-risk when it exceeds configured thresholds. Use when the user asks "what's my drawdown", "set up drawdown halt", "should I stop trading", "DD monitor", "max drawdown", "circuit breaker", or has a live trading process and wants automatic risk-off b
emos-bias-correctionApply EMOS / Platt scaling to correct systematic bias in a probabilistic forecast at inference time. Use when the user asks "fit EMOS", "apply Platt scaling", "calibrate my model output", "correct model bias", "my reliability diagram has a slope", "the model is over-confident", or has a calibration
kelly-sizerCompute optimal bet sizing on a prediction-market contract using the Kelly criterion. Use when the user asks "how much should I bet", "Kelly size this trade", "what's my optimal stake", "fractional Kelly", "is this position too big", or supplies a (win-probability, contract-price) pair and wants a r
maker-pricingCompute a maker quote that pays the maker fee tier (often zero) instead of the taker fee while still being likely to fill. Use when the user asks "what's the maker price", "quote inside the spread", "should I be a maker or taker", "what limit price should I set", "how do I minimize fees on this trad
market-scannerScan open prediction-market contracts for mispricings between your model's probability and the market mid, applying disagreement filters, side-specific edge thresholds, and extreme-price guards. Use when the user asks "scan markets", "find mispricings", "what should I bet on", "what are the opportun
pnl-attributionSlice realized P&L by side, price bucket, model agreement, cohort, and time to find where the strategy is making money vs leaking it. Use when the user asks "where am I losing money", "slice my P&L", "P&L attribution", "are the new filters earning", "is the YES side back to negative", "kpi report",
pre-flight-checklistRun the pre-flight safety checklist before flipping a trading bot from dry-run to live. Use when the user says "go live", "flip to live", "DRY_RUN false", "trade for real", "send it", "I'm about to run live", "pre-flight", "is the bot safe", or is about to launch the bot after any code change, model