OpenClaw Checkpoint - Personal AI Assistant Backup & Recovery (Github)

ReviewAudited by ClawScan on May 10, 2026.

Overview

This looks like a real backup-and-restore skill, but it needs review because it backs up private AI memory to GitHub, relies on external install scripts, and its docs understate what files and persistence it touches.

Before installing, review the GitHub installer/scripts, use a private repository, verify what files will be committed and excluded, understand all paths the skill can touch, and enable scheduled backups only if you want recurring background pushes.

Findings (6)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

A user may believe only the workspace directory is touched, while the skill can also create persistent scheduling configuration outside that directory.

Why it was flagged

The same security section discloses scheduler changes outside the workspace while also claiming no access outside the workspace, which materially understates the operating boundary.

Skill content
macOS: Creates a launchd plist at `~/Library/LaunchAgents/com.openclaw.checkpoint.plist` ... Linux: Adds a user-level cron job ... File access scope: The skill only reads and writes within `~/.openclaw/workspace`. It does not access files outside this directory.
Recommendation

Update the documentation to list every read/write path, scheduler change, and restore/delete effect explicitly, and require clear user confirmation for operations outside the workspace.

What this means

Installing this way requires trusting the current contents of the GitHub repository at install time.

Why it was flagged

The quick install fetches a mutable remote shell script and executes it; the executable scripts are not included in the reviewed package.

Skill content
curl -fsSL https://raw.githubusercontent.com/AnthonyFrancis/openclaw-checkpoint/main/scripts/install-openclaw-checkpoint.sh | bash
Recommendation

Review the remote installer before running it, prefer a pinned release or commit, and consider including the scripts in the packaged skill for review.

What this means

Private memories and agent context could be exposed if the repository is public, shared, or compromised, and restored memories may influence future assistant behavior.

Why it was flagged

The skill intentionally stores persistent assistant memory and identity data in a remote git repository.

Skill content
Automatically sync your OpenClaw agent's identity, memory, and configuration to GitHub. ... Your backup contains personal data: Agent identity and personality; Conversation history and memories; Personal notes and configurations
Recommendation

Use only a private repository, review exactly what will be committed, verify secret exclusions, and protect the GitHub account and keys used for backup.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

After scheduling, backups may continue running in the background and push updated assistant state to GitHub.

Why it was flagged

The skill can create recurring background backup behavior, but the documentation says it is opt-in and user-level.

Skill content
checkpoint-schedule sets up cron (Linux) or launchd (macOS) for auto-backups ... Auto-backup is opt-in only -- it is never enabled unless you explicitly run `checkpoint-schedule`.
Recommendation

Run scheduling only if you want recurring backups, and verify or disable it with checkpoint-status, checkpoint-stop, crontab -l, or the documented launchd plist.

What this means

A broad token or SSH key could allow more repository access than needed if mishandled.

Why it was flagged

GitHub authentication is expected for this backup integration, but it grants repository access that should be scoped carefully.

Skill content
Requirements ... SSH key or GitHub Personal Access Token ... A private GitHub repository for storing backups
Recommendation

Prefer SSH or a narrowly scoped token, use a private repository, and avoid sharing credentials with unrelated tools.

What this means

Running the command without understanding it could delete local backed-up agent files.

Why it was flagged

The static scan found a destructive delete command in documentation; no supplied artifact shows it is automatic, but it can remove an agents backup directory if run.

Skill content
rm -rf ~/.openclaw/workspace/agents
Recommendation

Only run reset or cleanup commands after reviewing what they delete and after confirming you have a current backup.