
Machine Checkpoint and Branch
Snapshot, schedule, restore, pause and branch smolvm machines, with cleanup scripts
- Repository
- smol-machines/smolvm
- Stars
- 6,901
- Licence
- Apache-2.0
- Files
- 13 files
What it does
It teaches an AI agent how to save the full running state of a smolvm virtual machine (disk and memory) as checkpoints, keep a bounded history of them on a timer, restore any saved version into a new machine, pause and resume machines, and fan a prepared machine out into child machines. The agent runs bundled shell scripts for these steps and checks the results before reporting success.
Great for
- Developers running long-lived or agent-driven smolvm machines who need rollback or periodic snapshots
- Teams that want to fan one prepared machine out into several identical workers
- People who need a restore to bring back running processes and RAM, not just files
Skip it if
- Moving an environment to another host as a file; the skill says that is a different packet (pack) and not covered here
- Simple persistence across a stop and start, which the skill says belongs to a separate dev-env packet
- Unattended production schedules without review: the scripts warn that checkpoints hold sensitive memory and disk contents, and the cleanup script can kill VM processes with --reap
Try asking
- “Take a checkpoint of my smolvm worker every 10 minutes and keep only the last 2”
- “Restore my smolvm machine from yesterday's checkpoint into a new machine called smolskill-old”
- “Pause my running machine, then resume it without losing its processes”
- “Branch my prepared machine into 4 workers, each starting from the same checkpointed state”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 13 original files, the Apache-2.0 licence and an attribution note.
branch-and-checkpoint.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip branch-and-checkpoint.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip branch-and-checkpoint.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Run scripts/preflight.sh first and stop if store_space_ok=no, since a full disk can publish nothing and leave machines hard to delete
- Set bounds on both --keep and --history, because kept checkpoints still retain earlier generations and the store grows otherwise
- Run cleanup.sh without --reap first; restarting a machine without --branchable loses anything in RAM or /tmp, so confirm with the user before doing it
What's inside
- disk-chain-compaction-qa.md
- README.md
- references/platforms.md
- references/scheduling.md
- references/traps.md
- scripts/branch-point.sh
- scripts/checkpoint.sh
- scripts/cleanup.sh
- scripts/create-source.sh
- scripts/preflight.sh
- scripts/restore.sh
- scripts/schedule.sh
- SKILL.md
Redistributed unmodified from smol-machines/smolvm under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-11-02. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.