Git Backup

WarnAudited by ClawScan on May 10, 2026.

Overview

This skill matches its stated backup purpose, but it can silently push broad private workspace data to GitHub and uses a powerful GitHub token in risky ways.

Only install or use this if you are comfortable sending your OpenClaw workspace to a private GitHub repository. Use a fine-grained token limited to the backup repo, review exactly what will be committed, strengthen .gitignore, and avoid silent scheduled backups unless you explicitly want them.

Findings (5)

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 GitHub token with broad repo access could expose or modify private repositories if mishandled, leaked through local config, or reused outside the intended backup repository.

Why it was flagged

The skill looks for GitHub tokens in local credential-like files and shell configuration, asks for a broad full-repo PAT, stores it in plaintext under ~/.credentials, and embeds it in the git remote URL.

Skill content
Check credential files ... ~/.credentials/github*.txt ... Check ~/.bashrc for embedded token ... Required permission: repo (full) ... echo "$TOKEN" > ~/.credentials/github-token.txt ... https://${TOKEN}@github.com/...
Recommendation

Use a fine-grained GitHub token limited to one private backup repository, avoid embedding the token in remote URLs, store credentials through a proper credential helper, and require explicit approval before reading local credential files.

What this means

Private memories, configuration, notes, and potentially sensitive workspace data may be copied to GitHub and retained in git history.

Why it was flagged

The skill stages all changes and explicitly includes memory, notes, installed skills, learnings, data, and configuration, which can contain sensitive user context or agent instructions.

Skill content
git add -A ... Always include: MEMORY.md, SOUL.md, AGENTS.md, TOOLS.md, PA_LIST.md ... memory/ ... skills/ ... .learnings/ ... data/ ... config/
Recommendation

Before enabling this, restrict the backup to specific safe paths, review .gitignore carefully, run git status/diff before each push, and verify the GitHub repository is private with appropriate retention controls.

What this means

The agent could continue pushing workspace changes in the background without the user noticing each backup or reviewing what changed.

Why it was flagged

The documented cron configuration creates ongoing autonomous backups every six hours and explicitly uses silent delivery.

Skill content
"schedule": "0 */6 * * *" ... "task": "Run git backup of the workspace to GitHub. Use the git-backup skill. Commit all changes and push. Report DONE or BLOCKED." ... "mode": "silent"
Recommendation

Do not enable the cron job unless you want ongoing background backups; prefer visible notifications, manual approval, or a dry-run review before each push.

What this means

Unintended files, secrets, or unwanted changes could be committed and pushed to the remote repository.

Why it was flagged

The regular backup flow stages every change in the workspace and pushes it, with no required file-by-file review or confirmation step.

Skill content
git add -A ... git commit -m "Auto backup $DATE" ... git push origin main
Recommendation

Add a mandatory git status/git diff review and require user confirmation before committing and pushing broad workspace changes.

What this means

A tampered .context file could run commands before the backup process.

Why it was flagged

Sourcing a local .context file is a shell execution mechanism; if that file is modified, its contents execute as shell code.

Skill content
CONTEXT_FILE="/opt/ocana/openclaw/workspace/skills/git-backup/.context" ... [ -f "$CONTEXT_FILE" ] && source "$CONTEXT_FILE"
Recommendation

Treat the .context file as trusted code, keep it user-controlled, and prefer parsing simple key-value configuration instead of using shell source.