Clustered Regularly Interspaced Palindromic Repeats (CRISPR)-CRISPR-associated (Cas) systems constitute adaptive immune mechanisms in prokaryotes that record past infections as spacer sequences within CRISPR arrays. While most discovery approaches rely on assembled genomes, metagenomic data offer access to uncultured microbial diversity but introduce computational challenges. The Metagenomic CRISPR Array Analysis Tool (MCAAT) detects CRISPR loci directly in succinct de Bruijn graphs without full assembly, yet lacks functionality to reconstruct nucleotide sequences. This thesis extends MCAAT by developing algorithms for spacer ordering and CRISPR sequence reconstruction from detected graph cycles. The ordering method derives pairwise constraints between cycles from read paths and applies priority-guided topological sorting to obtain biologically plausible spacer orders. Sequence reconstruction follows established CRISPR properties by identifying repeats through multiplicity and extension patterns, then extracting spacers from ordered cycles. Evaluation on 5095 reference loci from CRISPRCasdb demonstrates an average similarity of 82.63% between reconstructed and reference sequences after quality filtering, with most deviations attributable to local spacer misorderings rather than systematic failures. The approach achieves linear runtime complexity in the number of reads and modest memory overhead. Two prediction metrics quantify reconstruction reliability and enable filtering of ambiguous results. This work demonstrates that accurate CRISPR discovery is feasible directly from metagenomic graphs, reducing dependency on genome assembly and enabling broader exploration of CRISPR-Cas diversity in complex microbial communities.