golden_ticket: autonomous options agent on Alpaca

golden_ticket is an autonomous options trading agent that buys long single-leg calls and puts (30–60 DTE) on liquid US large caps through Alpaca's Trading API and has run unattended on a paper account since July 2026. AI logic: a 9:00 ET universe from Alpaca's tradable list; a scanner every 5–10 minutes that prices near-ATM contracts by exact symbol from Alpaca option snapshots and scores trend, breakout, IV rank, required move and liquidity; a screen that rejects crashes, spikes and moves that cannot pay the target; a buy engine that sends marketable limits anchored to the last real print. A learning loop attributes every closed trip per signal with Wilson intervals, scores execution, and runs a shadow ledger of untaken candidates; a 577-ticker Monte Carlo simulator is used only to reject ideas, never to endorse them. Risk gates: a 12-second exit stack (soft stop −8%, trail arm +12% with 3-point giveback and 6% lock, fade, stall, 3-day max hold, earnings exit), a GTC stop resting at Alpaca at −12%, and a risk governor with a 5% daily-loss halt, VIX-proxy breaker, portfolio heat and drawdown throttles. 25 risk parameters are range- and relationship-checked at boot; the engine refuses to start on an incoherent set. Alpaca infrastructure: Trading API for orders, positions and fills; Market Data API for bars, chains and snapshots; the official Alpaca CLI for broker-vs-engine reconciliation (npm run reconcile). 60 unit tests, a live smoke suite, pm2 with a launchd watchdog, no secrets in the repo. Record: 123 real round trips, 40.7% win rate, net negative, reported honestly. Hackathon paper account: PA35B89XVBYS (fresh, $100,000). Repo: github.com/salifhidone/golden_ticket