Set the technical direction for the AI-assisted tooling that implements, reviews, and validates engine changes and findings, design the checks that decide whether an agent-written change or generated result is fit to merge, and measure detection quality against benchmark applications with known vulnerabilities, raising the bar for what counts as a trustworthy result. Build and maintain the harness, agent instructions, and agentic skills that keep agent-written code trustworthy, including separate coding and reviewing agents, rules and tests the coding agents cannot change, independent review panels of agents from different model providers, and performance, API, and dependency gates, with risky changes escalated to human review.