A profile someone else built arrives in a particular emotional shape. It comes with a number attached, usually a good one, and with none of the failed versions that preceded it. You did not watch it being made, so you cannot see the forty variants that were tried and discarded, and the one you were handed is by definition the one that looked best at the end of that process.
That does not make it worthless. Plenty of shared profiles are perfectly reasonable. It means the burden of proof sits with you rather than with the number you were shown, and there are four checks that shift it. None of them take more than an afternoon and all of them happen before the profile places a single real order.
Import is the easy part, and the tab order tells you the rest
Autopilot has an import control sitting in the actions row at the top of the page, next to pause all, resume all, the emergency kill switch and the new profile button. That is how a profile built by somebody else gets into your account, and it takes one click, which is the whole problem in miniature.
Two tabs along is the Backtest tab, and the panel there tells you exactly what it is for. It replays historical signals against a profile with no real orders placed, it honors all filters, sizing, leverage caps and risk guards, and the panel describes itself as useful for validating before going live. That last phrase is not marketing, it is a sequence. Import, validate, paper, live. The number of people who do the first and fourth and skip the middle two is high, and it is a large share of why shared profiles have the reputation they do.
Read every setting before you run anything
Open the imported profile in the editor and write down what it contains, in your own notes rather than in your head. You are looking for four things, and the last one is the one people miss.
The signal source it follows. Profiles on this platform are frequently followers, named for the engine they consume rather than for logic of their own, which means importing one creates a dependency on a source. Check that you actually have access to that source and that it produces signals at the cadence the profile expects. A profile following a daily engine and a profile following a four hour engine behave nothing alike even with identical risk settings.
The instrument universe. Count the names. A profile listing forty instruments and a profile listing four are different products, and the wide one will demand far more simultaneous capital than the shared result implies.
The sizing rule and every guard. Position sizing, leverage cap, stops, trailing behaviour, partial take profit levels, and any risk guards with their values. Autopilot describes trailing stops and partial take profits as things a profile can carry, so a profile handed to you may have several exit mechanisms interacting, and you should be able to state what each one does before you enable it.
And then the fourth thing: anything you cannot see. If the editor does not display a setting, or displays it in a form you do not understand, that is a setting you are accepting blind. Write those down as a list and ask the person who built it. A builder who can explain each value has thought about them. A builder who cannot remember why the number is what it is has told you something useful about how it was chosen.
The lookback you were shown is not the one you should run
Every shared result carries an unstated window, and the window is where most of the flattery lives. The backtest panel takes a day count that runs backward from the moment you press the button, so you do not need to know what period the builder used. You can simply test several.
Run the profile at three or four different lookbacks on the same afternoon, with the same start equity throughout, and put the results side by side. Something short, something around six months, and the longest window the panel will give you. Then read the pattern rather than the best number.
A profile that holds up across every window is showing you a property of the strategy. A profile that looks excellent at one length and ordinary at the others is showing you a property of that length, and the length it looks excellent at is very often the one you were shown. That is not necessarily dishonesty. People run the window they ran and report it. It is still your money.
Use your own account size in the start equity field, not a round default and not whatever the builder used. A profile that works at fifty thousand dollars can be untradeable at five, because forty instruments and a concurrency ceiling mean positions too small to survive commissions. Work out your intended average position size, apply your venue's real round trip cost, and compare it to the average result per trade in the run. If friction is a large fraction of the average trade, the profile does not survive the trip to your account and no amount of validation changes that.
The universe test that catches a profile picked after the fact
This is the check that most reliably separates a strategy from a selection, and it takes one extra run.
Look at the trade list from your longest backtest and rank the instruments by contribution. If the profile lists forty names and three of them produced most of the result, you are probably not holding a strategy that works on forty instruments. You are holding a strategy that worked on three, wrapped in thirty seven that were included because including them made the aggregate look diversified.
So remove the top contributors from the universe, leave everything else untouched, and run it again. If the remaining thirty seven still produce a coherent result, the logic generalises and the top names were luck of the draw. If the result collapses, the profile is a bet on a handful of instruments and should be evaluated as one, including whether those instruments will keep behaving that way and whether you would have chosen them yourself in advance.
The same logic applies to time. Split your longest window in half and run each half. A profile that only works in one half is telling you which market it was built during.
The two weeks in paper mode that people skip
Everything above is simulation, and the platform is explicit about the limits of that on the same page. With paper mode off, Autopilot places real orders on your connected exchanges, execution prices may differ from signal prices because of market conditions and latency, and you are solely responsible for every trade. The stated route is to start in paper mode, verify the behaviour, then switch to live with small sizes. The header shows how many of your profiles are enabled and how many are in paper, and it is worth glancing at, because the common state is that nothing is being tested at all.
What paper mode catches that no backtest can is the gap between the profile you read in the editor and the profile that actually runs. Watch three things over a fortnight. Whether entries arrive at the frequency the backtest implied, because a large discrepancy usually means the signal source is not delivering what the replay assumed. Whether the exits behave as described, particularly if there are trailing stops and partial take profits interacting, since that is the area where a shared profile most often does something its own builder did not intend. And whether any guard ever binds, because a limit you have never seen take effect is a guess rather than a control.
Then go live small and keep it small for longer than feels necessary. The failure mode with an imported profile is not that it blows up on day one. It is that it works, you scale it up on the strength of a few good weeks, and the first genuinely bad market arrives when the position sizes have grown to match your confidence rather than your evidence. Find the emergency kill switch and the pause control before that day, and be clear with yourself that they stop new activity rather than undoing what has already happened.