Searching for new antimicrobials is a pressing issue to conquer the emergence of multidrug-resistant (MDR) bacteria and fungi. Antimicrobial peptides (AMPs) usually have antimicrobial mechanisms different from those of traditional antibiotics and bring new hope in the discovery of new antimicrobials. In addition to antimicrobial activity, stability and target selectivity are important concerns to decide whether an antimicrobial peptide can be applied in vivo. Here, we used a simple de novo designed peptide, pepD2, which contains only three kinds of amino acid residues (W, K, L), as an example to evaluate how the residues and modifications affect the antimicrobial activity against Acinetobacter baumannii, stability in plasma, and toxicity to human HEK293 cells. We found that pepI2 with a Leu→Ile substitution can decrease the minimum bactericidal concentrations (MBC) against A. baumannii by one half (4 μg/mL). A D-form peptide, pepdD2, in which the D-enantiomers replaced the L-enantiomers of the Lys(K) and Leu(L) residues, extended the peptide half-life in plasma by more than 12-fold. PepD3 is 3-residue shorter than pepD2. Decreasing peptide length did not affect antimicrobial activity but increased the IC50 to HEK293 cells, thus increased the selectivity index (SI) between A. baumannii and HEK293 cells from 4.7 to 8.5. The chain length increase of the N-terminal acyl group and the Lys→Arg substitution greatly enhanced the hemolytic activity, hence those modifications are not good for clinical application. Unlike colistin, the action mechanism of our peptides relies on negatively charged lipids rather than lipopolysaccharides. Therefore, not only gram-negative bacteria but also gram-positive bacteria can be killed by our peptides.