Sage — Mythic-based AI assistant for automating post-compromise actions in penetration testing

ToolsYesterday, 09:31
Sage — an AI assistant integrated into the Mythic v4 C2 framework as a chat interface, designed to automate operations for simulating adversary activity after gaining initial access to the infrastructure.
Sage converts operator requests into a sequence of tasks and commands for the relevant Mythic agents.
To work with context, Sage uses the BloodHound graph and the Engagement State Ledger, which stores information about the current context, completed actions, and achieved objectives.
Features:
Multi-agent architecture The main coordinating agent (Supervisor) manages six specialized agents. They interact with the Mythic infrastructure through active connections (callbacks) and work with tasks, credentials, files, and attack paths determined based on the BloodHound graph.
Tool preparation Sage checks whether the required executables are available in Mythic. If a required tool is missing, the assistant searches for it in the local tools directory and uploads it to the platform if necessary.
Technique and scenario library (TTPs) Sage includes a library of more than 200 playbooks. Each contains a technique description, an action sequence, and the pre-checks required before applying it.
Action control and confirmation In Supervised mode, operations require mandatory operator confirmation. This applies to creating and launching tasks, using agent capabilities, generating payloads, and adding credentials.
Operating modes Conversation — dialogue and analysis mode. Sage operates in read-only mode: it reviews the available context and helps the operator plan further actions without executing them. Supervised — mode for executing operations under operator control. Each action requires separate confirmation before being submitted for execution. An active Mythic agent connection (callback) is required to execute commands and launch tasks.
Vendors
Products