As offensive tooling becomes increasingly autonomous, the line between detection and prevention keeps moving. My current focus is building systems that learn the intent behind an attack rather than the signature.
MCP as the Control Plane
Instead of shipping another dashboard, I leaned on the Model Context Protocol as the connective tissue. Each tool becomes a composable capability that any MCP client can trigger with structured prompts.
Guardrails First
I implemented approval nodes that interrupt the flow whenever credentials, destructive payloads, or persistence actions are requested. Operators can attach notes, tweak parameters, and resume the chain safely.
Evidence Bundles by Default
Every executed capability produces:
- The raw command and stdout/stderr
- Sanitised artefacts (screenshots, PCAPs, memory dumps)
- A narrative summary ready to drop into the final report
This keeps the human in charge while the machine handles the mechanical work.