Model card: Galvanize 2 replaces this model in the cloud check; correct local-use claim; held-out and public benchmark results

Publish the FP32 source weights of the INT8 model and correct the model card

docs: fix MCP quick-start command (npx -y zn-gate mcp); update dataset/repo links

Initial release: Galvanize-60M agent-native prompt injection classifier

Initial release: Galvanize-60M agent-native prompt injection classifier

initial commit