Model Failover Guard

ReviewAudited by ClawScan on May 1, 2026.

Overview

This skill appears purpose-aligned, but it can continuously change OpenClaw's default model and restart the gateway, so users should review its settings before running it as a daemon.

This skill looks coherent for automatic model failover. Before installing or running it, review config.json, set excludedProviders if some models should never be used as fallback, run the script in once mode first, and only enable loop/systemd mode if you want it to keep changing the default model and restarting the gateway automatically.

Findings (4)

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.

Medium

#ASI02: Tool Misuse and Exploitation

What this means

The skill may change which model your agents use and restart the local gateway automatically, which can affect behavior, availability, and provider costs.

Why it was flagged

The script programmatically changes OpenClaw's default model and restarts the gateway. This is central to failover, but it is a high-impact action.

Skill content

set_primary_model(cfg, model_id)
save_openclaw_config(cfg)
subprocess.run(['openclaw', 'gateway', 'restart'], capture_output=True, text=True)

Recommendation

Run it once manually first, review the fallback/exclusion settings, and back up ~/.openclaw/openclaw.json before enabling loop or service mode.

Medium

#ASI03: Identity and Privilege Abuse

What this means

Installing and running this gives the guard authority to alter the local OpenClaw model configuration used by your agents.

Why it was flagged

The skill explicitly uses the user's local OpenClaw configuration authority to read the current model and modify the configured primary model.

Skill content

从用户 `~/.openclaw/openclaw.json` 读取当前默认主模型... 仅在触发条件满足时改 `agents.defaults.model.primary` 并重启 gateway

Recommendation

Confirm that automatic model switching is acceptable for your OpenClaw setup, and use a narrowly configured config.json with excluded providers where needed.

Medium

#ASI10: Rogue Agents

What this means

If started in loop or service mode, the guard can continue operating and making model changes until stopped.

Why it was flagged

Loop mode keeps the guard running indefinitely and repeatedly performs health checks and possible failover actions.

Skill content

while True:
        try:
            run_once()
        except Exception as e:
            log(f'run_once error: {e}')
        time.sleep(R['CHECK_INTERVAL_SEC'])

Recommendation

Only enable persistent mode if you want continuous failover, monitor ~/.openclaw/failover.log, and know how to stop the process or user service.

Low

#ASI04: Agentic Supply Chain Vulnerabilities

What this means

The skill will not function correctly unless the OpenClaw CLI is available, and users may not see that requirement from metadata alone.

Why it was flagged

The runtime depends on the OpenClaw CLI, while the registry metadata declares no required binaries. This appears expected for the purpose, but the dependency is under-declared.

Skill content

cmd = [
        'openclaw', 'agent', '--agent', 'main',
        '--message', '只回复OK', '--json', '--timeout', str(R['TEST_TIMEOUT_SEC'])
    ]

Recommendation

Verify the installed source and ensure the OpenClaw CLI is installed and trusted before running the script.