Security audit
Security checks for vulnerabilities and agentic risk
This skill performs a narrow FastRouter setup for OpenClaw, with expected config changes and no evidence of hidden exfiltration or malicious behavior.
Install this if you want Codex to configure FastRouter in OpenClaw. Before running it, check whether you already have custom FastRouter settings because the skill refreshes that provider block, and only approve the gateway restart when a brief interruption is acceptable.
The skill asks for a user API key but then fetches the model list without using authenticated headers, so it may retrieve an incomplete or generic catalog instead of the user's actual available models. This can silently misconfigure OpenClaw, leading to broken model entries, failed requests, or incorrect assumptions about what the account can access.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
Use web_fetch to get the model list:
web_fetch url="https://api.fastrouter.ai/v1/models" extractMode="text"
The skill instructs overwriting the existing models.providers.fastrouter configuration with a fresh model list, which can destroy prior custom settings, manually curated models, or nondefault provider options. Because this write targets a user config file, the lack of a prominent warning and confirmation increases the risk of unintended configuration loss.
Restarting the OpenClaw gateway can interrupt active sessions or temporarily break dependent workflows, yet the skill does not clearly explain that operational impact before asking for approval. User approval alone is not sufficient if the consequences of that approval are not made explicit.
Line L008 says the setup uses only built-in tools and requires no script execution. Later, L082-L084 instruct running openclaw gateway restart, which is command execution and therefore contradicts the earlier claim about how the skill operates.