2026-03-20

A dry-run checklist before you enable live routing

Repeatable gates so your first session is boring on purpose.

Live capital should be boring. Dry-run exists so you observe timing, gate cadence, and logging without paying tuition. This checklist is generic enough for Polymarket-class CLOB workflows; your artifact README always wins if it conflicts.

Environment and secrets

[ ] You are on the intended machine (or container) you will use live, not a laptop you will shut down mid-session.
[ ] Env files match the profile you think you are running (no accidental prod keys in a dev path).
[ ] Wallet and API credentials are scoped the way the runbook describes: one writer, no duplicate bots on the same nonce space.

Clock and network

[ ] System clock verified (NTP healthy). Skew shows up as false confidence near settlement.
[ ] Outbound connectivity to required feeds tested once in the last 24 hours, not assumed from last week.

Dry-run duration and scope

[ ] At least two full sessions in dry-run, or the minimum your docs require, whichever is longer.
[ ] You have reviewed audit output: gate reasons, order attempts, and any “would have filled” lines make sense for the book state you saw.

Operational readiness

[ ] Shutdown sequence is written and you have rehearsed it once (keyboard shortcuts or systemd stop, not improvisation).
[ ] Someone knows you are running live and how to reach you if the venue halts trading.

Promotion to live

Only after the above, flip live routing in the single place your kit documents. If anything in dry-run felt “noisy” or inconsistent, fix that first. Strategy tweaks are cheaper than reconciling a bad first session.

Where CLOB Forge fits

Licensed kits are documented for artifact install, profiles, and deployment after purchase. Orientation for buyers lives in the Polymarket bots guide; the catalog is on products. For a concrete starting SKU, many operators compare the 5m scalp kit and the 15m scalp kit before adding mean reversion or copy-style workflows.