by 安全扫描
OpenClaw
可疑
medium confidenceThe skill's instructions mostly match its stated purpose (fixing model routing) but it reads/writes hidden OpenClaw config files and hardcodes an external provider endpoint while the registry metadata did not declare required config paths or a trusted source — proceed with caution and review/back up configs first.
评估建议
This skill performs direct edits to OpenClaw configuration and restarts the gateway while pointing the provider to a specific external endpoint (coding.dashscope.aliyuncs.com). Before running it: 1) Back up $HOME/.openclaw/openclaw.json and any agents/*.json files; 2) Manually review the configs it will change and verify the hardcoded endpoint is correct and trusted for your environment; 3) Confirm you have the 'openclaw' CLI and permission to restart the gateway; 4) Prefer running the audit ste...详细分析 ▾
ℹ 用途与能力
The skill's described goal (diagnose/fix model routing) aligns with the actions in SKILL.md (audit config, update provider URL, remove agent override, restart gateway). However, the registry metadata claims no required config paths or credentials while the runtime instructions explicitly read and modify $HOME/.openclaw/openclaw.json and $HOME/.openclaw/agents/main/agent/models.json. That metadata omission is an inconsistency that reduces transparency.
⚠ 指令范围
The instructions tell the agent to read and overwrite user config files, delete an agent override file, and restart the OpenClaw gateway. They also hardcode a specific external endpoint (https://coding.dashscope.aliyuncs.com/v1). Modifying provider endpoints can redirect model traffic to a third-party host; deleting agent files is destructive if done without backup. The SKILL.md does not include any safety checks (backup, confirmation, or validation of the endpoint) before making changes.
✓ 安装机制
Instruction-only skill with no install steps or code files — lowest-risk install surface. There is nothing being downloaded or installed by the skill itself.
⚠ 凭证需求
The skill declares no environment variables or credentials, and it doesn't attempt to read secrets explicitly, which is consistent. However, it requires write access to OpenClaw config files and the ability to run the 'openclaw' CLI to restart the gateway. More importantly, it changes the model provider endpoint to a hardcoded third-party URL; that is a high-impact network change and should be justified and verified before applying. The skill does not request or declare this trust requirement.
ℹ 持久化与权限
always is false and the skill is not force-included. Still, the instructions modify persistent agent/global config and restart the gateway — an action requiring filesystem and service control privileges. The skill does not attempt to persist itself beyond those changes, nor does it modify other skills' configs, but the changes it makes are persistent and potentially disruptive.
安全有层次，运行前请审查代码。
运行时依赖
无特殊依赖
版本
latestv1.0.02026/3/1
- Initial release of model-route-guard. - Diagnose and fix model routing conflicts for Openclaw. - Audit and resolve duplicate provider overrides and endpoint mismatches. - Ensures primary model uses correct provider endpoint (coding.dashscope.aliyuncs.com). - Provides step-by-step workflow for route auditing, conflict fixing, and verification. - Safeguards privacy by not exposing API keys in output.
● 可疑
安装命令
点击复制官方npx clawhub@latest install model-route-guard
镜像加速npx clawhub@latest install model-route-guard --registry https://cn.longxiaskill.com