Curated public work

Projects with status and scope.

Each entry links to public source and identifies whether it is maintained work, a reference design, or archived.

maintained

Llama Hero

Podman Quadlet configuration for a self-hosted llama.cpp inference server, including model metadata and a VRAM-budget validation tool.

reference

Helm chart source for packaging non-secret OpenCode configuration, role-specific agents, subagents, and specialized skills for a shared server.

archived

An archived multi-agent coordination experiment. Its documented decommissioning captures maintenance and prompt-injection trade-offs.