How to use the Local Database Backups agent skill (backup-setup)

What an agent skill is and what this one does

An agent skill is a folder containing instructions (a SKILL.md file, plus any supporting files) that an AI agent reads when a request matches the skill's description. Once installed, you do not run it yourself; you ask for the task in ordinary words and the agent follows the instructions.

This skill covers three jobs: setting up scheduled local snapshots of the claude-mem database at ~/.claude-mem/claude-mem.db, checking their status, and restoring a snapshot. Each snapshot is a standalone SQLite file stored under ~/.claude-mem/backups/auto/, and the newest copies are kept up to a count you choose. Snapshots stay on your own machine.

Before you start: what the skill needs

The skill depends on claude-mem being installed with its npx command-line tool and a running worker. It also needs Node.js, because the setup step runs a Node script, and SQLite support through the claude-mem worker.

The skill runs scripts, and it does not use the network, according to its source.

• claude-mem installed, with its npx command-line tool available
• A running claude-mem worker (the skill can start it with npx claude-mem start)
• Node.js installed
• SQLite support through the claude-mem worker

Install the backup-setup skill

Download the skill ZIP from this page and unzip it. Then put the resulting folder in one of two places: ~/.claude/skills/ to make it available across your projects, or a project's .claude/skills/ folder to limit it to that project.

If you use the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into the project's .claude/skills/ folder
• Claude apps: upload the ZIP under Settings → Capabilities → Skills

Ask for it with natural requests

The skill triggers when you use phrases like the ones in its description. You do not need to name the skill or use a command. The skill also accepts an argument of setup, restore or status, and its routing maps an empty request to setup.

• "Set up backups" or "back up my memories": checks status, then enables snapshots
• "Backup status": reports the last snapshot time, snapshot count and any last error
• "Restore my database": lists snapshots and restores the one you choose

What the agent does when you enable backups

For setup, the agent first runs the status check. If backups are not yet enabled, it asks you for an interval in hours (default 24, range 1 to 168) and how many snapshots to keep (default 7, range 1 to 100). It then writes three backup settings into ~/.claude-mem/settings.json, keeps the file's mode at 0600, and preserves your other settings.

After writing the settings, the agent restarts the worker and checks status again. It reports backups as on only when the status output says so. The first automatic snapshot lands about five minutes after the worker starts. You can take one immediately with npx claude-mem backup run.

How to restore a snapshot safely

For a restore, the agent lists available snapshots, then restores the file you pick. The restore stages the snapshot and restarts the worker. The new worker saves the database it replaced as claude-mem-<timestamp>-pre-restore.db in the same folder. You can restore that file to undo the change, and it counts toward your retain count.

After a restore, backup status shows a "Last restore" line that names the saved copy on success, or the reason for failure. If it fails, the worker keeps the database it had.

Examples of requests and what happens

Each example below shows a request you might type and the steps the skill follows, based on its instructions.

• "Set up backups, every 12 hours, keep 10": the agent checks status, writes the interval and retain count, restarts the worker and confirms the schedule.
• "Backup status": the agent runs the status check and reports the snapshot count, last snapshot time and any error, then stops.
• "Restore my database from yesterday": the agent lists snapshots and restores the one you choose, then reports the restore outcome.

Safety notes: scripts, settings and cloud sync

The skill runs scripts, including a Node script that edits your settings file. The skill is written to avoid printing that file's contents, because it can hold credentials. Keep the file's permissions at 0600.

Restores need care when cloud sync is on. The restored database resumes syncing from the snapshot's position and can pull back changes that reached the cloud after the snapshot was taken. Snapshots are local only, so they do not protect you from losing data stored off the machine.

• The skill runs scripts and does not use the network.
• Snapshots are local only; there are no off-machine or cloud backups.
• Restoring with cloud sync on can bring back changes made after the snapshot.

Tips for getting reliable results

Check status before and after changing anything. If the status output reports an unknown backup command or an HTTP 404, the installed claude-mem version may not include backup support yet; the skill says to run npx claude-mem update in that case. Other errors should be read and not ignored, since the skill will not turn backups on after them.

• Run backup status first to confirm the worker is up.
• Keep retain count high enough to cover the window in which you would notice a problem.
• Use the pre-restore file if a restore turns out to be wrong.
• If the worker is not running, the skill will try npx claude-mem start and retry.