The visual target
I wanted flat cuts that crossed an edge and reached the adjacent faces, similar to a deliberate sculpting stroke. Rounded dents and uniformly noisy silhouettes did not match that target.
The experiment began in 3ds Max and moved to Blender Geometry Nodes. The agent helped construct and revise the procedural approaches while I evaluated the shape of the damage.
Several approaches, different failures
Uniform subdivision made the work expensive. Adaptive bands introduced seams. Volume remeshing changed the construction again. A later approach used cube-shaped cutters to move closer to planar chips.
The records include topology problems, stale evaluated-mesh readings and two crashes after subdivision was increased without sufficient safeguards. Repeated numbers were sometimes evidence of cached geometry being read again, not proof that a parameter had no effect.
Current limit
I do not have a finished replacement for a sculpting pass from this experiment. Topology, UV handling, stability and the visual quality of the cuts still need to be demonstrated together.
It is a useful record of why matching an artistic operation requires more than reproducing its rough technical description.