v1.2.0: Autonomous Red Teaming Update

Release v1.2.0: Autonomous Red Teaming Update

This release transforms Gideon from a defensive intelligence analyst into a Dual-Mode Autonomous Security Platform, introducing safe, sandboxed offensive capabilities alongside our existing defensive guardrails.

What's New

🛡️ Dual-Mode Architecture

• Red Team Mode: Added the /redteam command to unlock authorized autonomous exploitation with strict engagement scoping limits.
• Dynamic Guardrails: NeMo Guardrails now dynamically unload when transitioning to Red Team mode to permit offensive workflows.

⚡ Action Engine & Sandboxing

• Introduced Docker-based Execution Sandboxes for all external tools executed by the LLM (like nmap, nuclei, sqlmap, metasploit) to guarantee host safety.
• Built a command orchestration engine that autonomously parses stdout/stderr vulnerability findings and injects them directly into the Neo4j intelligence graph.

📡 C2 Framework Interoperability

• Added a unified C2 interface adapter system for persistent session management.
• Implemented native gRPC support for the Sliver C2 framework, enabling Gideon to locally list sessions, deploy stagers, and execute commands through active implant beacons.

🛠️ Expanded Offensive Skills

• Weaponization: Added an autonomous payload generator leveraging msfvenom to build, encode, and obfuscate payload loaders for EDR evasion.
• Post-Exploitation: Integrated Active Directory network mapping into the graph database, powering an intelligent lateral movement decision engine for privilege escalation tracking.

🔐 OpenClaw Sentinel Integration

• Released fully updated configuration models and documentation for OpenClaw Sentinel (acting as an AI-powered WAF and sidecar).

🐛 Bug Fixes & Chore

• Resolved all pre-existing TypeScript compilation errors (@types) and implicit any traces across the graph, engine, openclaw, and mcp subsystems.
• Synchronized missing dependencies in package.json and updated the docs/ sidebar navigation.

Full Changelog: v1.1.1...v1.2.0