I assume you are talking about analytical HLPC.

The answer depends on whether you are doing a calibration curve or just wish to identify a peak. It also depends on the strength of the chromophore, the wavelength that you are using, the sensitivity of the instrument, the cleanliness of the baseline and the flow rate of the method that you use.

Generally one wants a signal to noise (S/N) ratio greater than 5:1. At that limit it is possibly for an Agilent detector to reliably detect as little as 1 ng, but that is pushing it. It is typical to make up a sample around 1-5 mg/mL then inject 1-10 microliters.

Without a calibration curve, do not confuse % AUC (area under curve) purity with mol % purity. Say you have an HPLC sample with two compounds, A & B, with 91% and 9% AUC respectively but the extinction coefficient of A is 10 times that of B then the mol % purity of the sample is 50:50. If you do a calibration curve then change the gradient or the flow rate you will have to redo the calibration curve.