A robust, high-throughput method has been developed to screen one-bead–one-compound peptide libraries to systematically profile the sequence specificity of protein kinases. Its ability to provide individual sequences of the preferred substrates permits the identification of sequence contextual effects and nonpermissive residues. Application of the library method to kinases Pim1, MKK6, and Csk revealed that Pim1 and Csk are highly active toward peptide substrates and recognize specific sequence motifs, whereas MKK6 has little activity or sequence selectivity against peptide substrates. Pim1 recognizes peptide substrates of the consensus RXR(H/R)X(S/T); it accepts essentially any amino acid at the S/T–2 and S/T+1 positions, but strongly disfavors acidic residues (Asp or Glu) at the S/T–2 position and a proline residue at the S/T+1 position. The selected Csk substrates show strong sequence covariance and fall into two classes with the consensus sequences of (D/E)EPIYϕXϕ and (D/E)(E/D)S(E/D/I)YϕXϕ (where X is any amino acid and ϕ is a hydrophobic amino acid). Database searches and in vitro kinase assays identified phosphatase PTP-PEST as a Pim1 substrate and phosphatase SHP-1 as a potential Csk substrate. Our results demonstrate that the sequence specificity of protein kinases is defined not only by favorable interactions between permissive residue(s) on the substrate and their cognate binding site(s) on the kinase but also by repulsive interactions between the kinase and nonpermissive residue(s).