ForgeyClap Forge playbook — Prediction / sports-data
Hard rules
- Never claim certainty. Every output carries a confidence/risk label.
- No automatic real-money betting / no auto-stake code path.
- Backtest before trusting any model; show sample size.
- Verify data-source quality; track all assumptions.
- Telegram token in env only; one message format, tested.
Team (conditional)
Lead: mle-reviewer + planner. Specialists: mle-reviewer, python-reviewer, silent-failure-hunter (data gaps look like clean zeros), database-reviewer (data store).
Skills / commands / MCP
systematic-debugging, /test-coverage. For delivery, defer to forge-n8n or a Telegram layer (Telegram bots fold in here).
Fan-out & flow
L3. Parallel: data ingestion ∥ model/stat logic ∥ delivery formatting. Serial: ingest → backtest → value/odds calc → uncertainty labeling → delivery.
Domain gates
Data sources + quality verified; statistical logic reviewed; backtesting done; odds/value checked; confidence/risk labels on every output; assumptions documented; data freshness check.
Ship-readiness (unique)
Backtest results shown with sample size; every output labeled with confidence/risk; assumptions + data-quality notes attached; no auto-bet path; Telegram message format tested. The ship-readiness prediction + Telegram checklists are advisory; optionally run codex-reviewer (Codex) on important code — not a blocker.