Model Failover Doctor
Security checks across malware telemetry and agentic risk
Overview
The skill mostly matches its stated OpenClaw failover-repair purpose, but its auto-fix can permanently rewrite model routing to a hard-coded provider, so it needs review before use.
Do not run --fix --restart blindly. First run the diagnostic and --dry-run, check whether any pool entry would be remapped to `custom-llmapi-lovbrowser-com`, and only proceed if that provider is intentionally configured and trusted in your OpenClaw setup.
VirusTotal
63/63 vendors flagged this skill as clean.
Risk analysis
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.
Running the auto-fix could send future OpenClaw model requests to an unexpected provider or break routing if that provider is not valid for the user.
The P-1 auto-fix can rewrite invalid model-pool entries to a hard-coded provider rather than only choosing among confirmed user-configured providers. This can change where future model traffic is routed.
lb = 'custom-llmapi-lovbrowser-com' ... # 3) 最后兜底：挂到 lovbrowser ... new_val = f'{lb}/{model_path}' ... pools[pool_name][field] = new_valUse --dry-run first, manually verify every provider change, and update the repair logic to choose only explicitly configured providers with user approval.
A mistaken repair could affect model selection across OpenClaw, not just the current failing request.
The skill explicitly edits global OpenClaw routing and session-state files. This is purpose-aligned, but changes can affect later sessions and subagents.
修复范围明确，只触碰与 failover 相关的三个文件： - `message-injector/index.ts` - `.lib/pools.json` - `.lib/session_model_state.json`
Keep backups, review the dry-run diff, and test with a non-critical session before restarting the gateway.
The tool can inspect local provider names and model configuration, though the visible code does not show credential exfiltration.
The tool reads local OpenClaw model-provider configuration to validate provider names. That access is relevant to the purpose, but it is not declared in the registry requirements.
OPENCLAW_JSON = BASE / 'openclaw.json' ... return set(cfg.get('models', {}).get('providers', {}).keys())Run it locally only from a trusted source, and avoid sharing generated output if provider names or model configuration are sensitive.
Users have less external context for who maintains the tool or where to verify its source.
The skill has limited provenance information. That matters more because it includes code intended to edit OpenClaw runtime files.
Source: unknown Homepage: none
Inspect the installed script before running repair mode, and prefer a reviewed or pinned source for production OpenClaw environments.