Why a broad suite of security skills makes focused research hard to repeat.
Interactive agents can find vulnerabilities. The harder problem is making security research repeatable across sessions and targets.
This series documents the controls behind that transition, including stable research roles, isolated contexts, durable state, and independent triage.
It continues from What’s Left When the World Becomes Agentic? .
Why a broad suite of security skills makes focused research hard to repeat.
Why leader-centric agent swarms do not preserve distinct security-research techniques.
Why security findings need a fresh context with the opposite mandate.
How isolated coding agents, Git coordination, and separate post-processing became an execution engine for continuous security research.