I developed a computational micro-grammar for a bounded cluster of Indus inscriptions using the Sigilith Structural Engine. Rather than attempting decipherment, the project models recurring structural behaviour through normalized symbols, transition rules and finite-state constraints. The resulting CORE + STROKE-GROUP + TAIL architecture can classify legal and illegal sequences, calculate structural stability, detect collapse points and generate or predict structurally valid sequences. The work demonstrates how formal computational methods can investigate an undeciphered symbolic system without making semantic claims.