It is not possible to prove deletion larger than the average read length from e.g. observing a coverage of 0 in a given location, because the lack of coverage might be due to random factors, in case you had that in mind. If you use samtools mpileup, and do not turn indel calling off (e.g. with -I), then indels should be called and be contained in the VCF file. A short deletion can be identified in the ALT column, e.g. as described here or simply by comparing the ALT and REF column.