
from vibe-check10
Tooling and workflows to build, convert, upload, and run Frostbite/Cauldron ML models on Solana, including deterministic account lifecycle and on-chain inferenc
Cauldron (Frostbite ModelKit) provides CLI-driven workflows to initialize model projects, convert and pack weights, build guest binaries, create seeded deterministic Solana accounts, upload model weights, and run on-chain inference. It codifies the end-to-end process for deploying lightweight ML models to a Frostbite program on Solana, with helper scripts and TUI guidance for both wizard and manual flows.
Use this skill when you need to prepare models for on-chain inference or test deployments on Solana devnet. Typical scenarios: converting local model weights into a packed format, building guest runtimes, creating seeded accounts for predictable addresses, uploading weights, and invoking the program for test inferences. Commands include cauldron init, convert, build-guest, accounts init/create, upload, and invoke.
This skill is most useful for agents with shell/CLI access and Solana tooling support (agents that can run cauldron commands or orchestrate Solana RPC calls).
This skill has not been reviewed by our automated audit pipeline yet.