Natural products (NPs) are rich source of drug discovery and some of them act by covalently binding to the targets. Recently, targeted covalent natural product (TCNP) design has gained considerable attention since this approach offers significant benefits in improving biological efficacy and decreasing off-target side effect. However, most of the known TCNPs were discovered by chance. Rational approaches for systematic screen of TCNPs are much needed. Here we developed a combined computational and experimental approach to carry out herb-based screening to identify TCNPs against proper cysteine residues in the target proteins. The herb-based TCNP screening approach (HB-TCNP) starts from druggable pocket and cysteine residue prediction, followed by virtual screening of covalent NP database and herb-based mapping to identify candidate herbs for experimental validation. Herbs with time-dependent activity are selected and their NPs are experimentally tested to further screen covalent NPs. We have successfully applied HB-TCNP to screen anti-PLK1 herbs and NPs with high efficacy. Cys67 and Cys133 in the ATP binding pocket of PLK1 were used in the search. Five herbs were tested and exhibited PLK1 inhibition activity to some extend, among which Scutellaria baicalensis showed the most potent activity with time-dependency. Further experimental studies showed that the main active compounds, baicalein and baicalin, in Scutellaria baicalensis covalently bind PLK1 through Cys133. Our study provided an efficient way to rationally design TCNPs and make better use of herb medicines. The Cys133 residue in PLK1 serves as a novel covalent site for further drug discovery against PLK1.