Elegant Config Guardian

ReviewAudited by ClawScan on May 10, 2026.

Overview

The skill is coherent for guarded OpenClaw config edits, but its rollback protection has gaps and it runs a caller-supplied shell command.

Review or fix the rollback logic before relying on this for production config changes. Only pass trusted patch commands, use --require-ack, verify that the gateway actually restarted with the new config, and keep the backup path handy for manual recovery.

Findings (3)

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 failed or partially failed config patch could leave OpenClaw configuration changed while the user believes automatic rollback protected them.

Why it was flagged

This safety claim is stronger than the implementation: the script uses an unguarded apply step under `set -e`, and some failures can exit before rollback logic runs.

Skill content
Use `scripts/safe_apply.sh` to enforce: backup → apply → restart → health check → optional ack wait → rollback on failure.
Recommendation

Add explicit error handling or a trap that restores the backup on any apply/restart/health failure, or document that rollback only covers specific health-check and ack-timeout failures.

What this means

The script may report success even when the intended runtime change was not applied, leaving a risky config change on disk for a later restart.

Why it was flagged

Gateway restart failures are explicitly ignored, so a later status check may not prove that the new configuration was actually loaded.

Skill content
openclaw gateway restart >/dev/null 2>&1 || true
Recommendation

Do not suppress restart errors; fail closed, roll back on restart failure, and verify that the running gateway actually loaded the new config.

What this means

If an unsafe command is supplied, it can run arbitrary local commands under the user's account before rollback checks occur.

Why it was flagged

The script accepts a caller-supplied command and executes it through shell eval. This is disclosed and central to applying config patches, but it is not constrained to the target config file.

Skill content
--apply-cmd) APPLY_CMD="$2"; shift 2 ;; ... eval "$APPLY_CMD"
Recommendation

Use only reviewed, trusted patch commands; avoid building --apply-cmd from untrusted text, and consider replacing eval with a safer, constrained command interface.