The flagged credential flows send tokens to their respective AI providers for usage reporting or configured captioning. The metadata and API-mutation indicators resolve to tests, and the inspected process execution performs a bounded DNS lookup.
Static reason
High-risk behavior combination matched malicious policy.; previous stored version diff introduced dangerous source
Trigger
Usage requests occur when the usage command runs; provider captioning occurs during the relevant media workflow when a provider key is configured.
Impact
The inspected behavior supports requested package features and did not show unrelated collection, credential forwarding, or harmful automatic actions.
Mechanism
The CLI contacts the selected provider with its own key or harness credential; a capture helper runs a local DNS lookup in a child process.
Network endpoints5
api.anthropic.com/api/oauth/usagechatgpt.com/backend-api/wham/usagecli-chat-proxy.grok.com/v1/billing?format=creditscli-chat-proxy.grok.com/v1/settingsopenrouter.ai/api/v1/chat/completions