Security and data handling

​

The control-plane boundary

• Sage may use Mythic and BloodHound control-plane data: callbacks, tasks, files, credentials, graph facts.
• All target-facing interaction executes as Mythic tasks on live payload callbacks. Sage never connects directly to target LDAP, SMB, Kerberos, WinRM, RPC, HTTP, or equivalent services, and Sage-local attack artifacts are not compromise proof.

​

Evidence rules

• Runtime effects and objective completion require Mythic task output or artifacts, Mythic credential-store state, or BloodHound facts derived from payload-collected artifacts.
• Self-reported success is not proof. The engagement ledger records a hop as achieved only with a valid proof envelope.

​

Data handling

• Do not put secrets in scripts, skills, logs, or public artifacts. read_credentials can place raw secrets in model context and traces; use it deliberately.
• Retained identity and membership observations are shared across assisted chats in the same Mythic operation engagement and persist across Sage restarts. /report shows what has been retained; a reported identity or host membership does not prove current authentication or access.
• The optional Slack hook emits only a fixed generic change notice: never operation, finding, host, user, domain, credential, task, or file values. Full findings stay in native Mythic chat inside your trusted deployment.