Are conducting propensity score matching to compare the effect of treatment in a balanced sample, however one of the factors associated with the assignment of treatment is previous levels of the outcome variable. All discussions I can find are on pitfalls when matching with the purpose of performing a diff-in-diff, which I am not. Right now i will perform simple cross-sectional regressions to estimate ATT. What is the problem of matching on the outcome variable pre-treatment when performing matching (PSM)?
The bigger question is what makes propensity scores a good idea in this setting? For a full discussion see https://hbiostat.org/bbr/propensity.html.
When you have a true baseline version $Y_0$ of the outcome variable, that one variable often encapsulates most of the confounding that's going on, and also explains a lot of the variation in Y. It is necessary to adjust for it against Y, not just how $Y_0$ affects the treatment. Another reason not to use propensity scores. Covariate adjustment is required.
Since you have non-randomized, observational data, there are a few situations where adjusting for $Y_0$ can lead to bias so beware. But usually it is the most important variable to adjust for.