The flagged credential behavior is part of user-invoked key rotation, where the report contains key names and paths rather than secret values. The prompt hook is opt-in, and CLI telemetry is gated on explicit enablement; no malicious behavior was identified in the inspected paths.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
Key rotation requires a user-run `gad keys` command; prompt suggestions require opt-in; telemetry requires explicit enablement.
Impact
The inspected source does not show automatic collection of secret values or credential exfiltration.
Mechanism
Enrollment reports environment-file key names and paths to Great Arrow, while storing device keys in the local keychain. The prompt hook sends the prompt text to Great Arrow using the user’s Great Arrow token.
Evidence
dist/index.jsdist/hooks/gad-suggest-tools.mjs
Network endpoints3
www.greatarrow.ai/api/mcp/tools-suggest/api/key-rotation/devices/api/telemetry