nx_stl_adversarial_test.nx

nx_stl_adversarial_test.nx -- Phase A1 of S-class hardening per docs/NISHI_3D_PRINT_SCLASS_HONEST_AUDIT_2026_05_20.md. Operator question: "are we really at s class or still theoretical bullshit, aka will it work like an ak47 in the worst of the worst" Honest answer: not S-class yet. This smoke is the FIRST STONE of adversarial hardening -- proves the binary STL parser gracefully rejects 8 classes of malformed input without crashing or returning false-OK verdicts. What this proves: parser DOES NOT crash on adversarial input + DOES return non-OK verdict on every malformed case + DOES detect class of corruption correctly (TOO_SHORT vs TRUNCATED vs BAD_FLOAT vs ASCII_REFUSED). What this does NOT prove: the parser is bug-free for ALL malformed inputs. Future work: fuzz with longer / randomized byte streams. expect_exit: 0 license_tier: ORIGINAL