Extract the rules, not the sales pitch
Upload the PDF and ask TRUE to identify the asset universe, entry condition, exit condition, position size, schedule and capital limit. Ask it to quote the relevant passage or identify the page for each rule.
Try: “Check whether this strategy is specific enough to implement. Extract its rules, list contradictions and mark anything the document leaves undefined. Do not fill in missing trading rules without asking me.”
Make vague instructions measurable
“Buy the dip” needs a reference price and a threshold. “Take profit quickly” needs an exit condition. “Use a small position” needs an amount or percentage. A cooldown also matters if a condition can remain true for hours.
For a DCA document, write down total budget, amount per purchase, interval and number of purchases. A $600 budget spread over twelve weekly buys means $50 each week, before costs. Check that the document and the proposed configuration use the same units.
Test the logic with examples
Ask what happens after a missed purchase, an unavailable market, a sudden price gap or insufficient funds. Look for conflicting rules, such as buying weekly while also stopping all purchases below a price threshold.
Separate backtest claims from the rules themselves. Check whether historical results include fees, slippage and a clear date range. A strategy can be implementable even when its claimed performance has not been demonstrated.
Create a draft and review it
Ask TRUE to map the agreed rules to a supported agent strategy. Through MCP, an assistant can create an agent draft, but funding and activation remain in the TRUE app. A PDF upload or a draft creation does not itself authorize trading.
Before signing, compare the draft with your extracted rule list. Review the first execution afterward. If you use a Jupiter recurring order instead of a TRUE agent, track and cancel that schedule separately; the two are different forms of automation.