smolvm

# smolvm skill packets

> Tested procedures, one task each, with their scripts. The flat pages in this directory are
> the reference for each feature.

- [branch-and-checkpoint](branch-and-checkpoint/SKILL.md): scheduled checkpoints with history, restores, pause and resume, and branches with their branch point kept
- [credentials](credentials/SKILL.md): an API key a workload can use but never read, substituted on the way out
- [dev-env](dev-env/SKILL.md): a machine you come back to, and what survives a stop and start
- [docker-in-machine](docker-in-machine/SKILL.md): a Docker daemon inside a machine, for tools that call Docker themselves
- [gpu-cuda](gpu-cuda/SKILL.md): CUDA Driver API calls remoted to the host NVIDIA GPU
- [install](install/SKILL.md): installing smolvm and proving the host can boot a microVM
- [local-api](local-api/SKILL.md): driving the machine lifecycle over the local HTTP API
- [pack](pack/SKILL.md): packing an image or a machine into a portable artifact
- [teardown](teardown/SKILL.md): what state exists, where it lives, and how to leave nothing running
- [throwaway-machine](throwaway-machine/SKILL.md): running untrusted code in a throwaway machine, with egress granted one host at a time

More agent context in smol-machines/smolvm

One other file this repository gives its agents.