Coverage-Driven Sustained Testing (CDST): Agentic Workflows
I wrote up CDST, a graph-oriented model for open-ended agentic workflows. Instead of agents chewing through a predefined task list, work is generated from a Neo4j coverage graph that tracks unexplored state dimensions (tested for XSS but not SSRF, authenticated but not anonymous, and so on). Discoveries during execution spawn new nodes, the graph picks the next gap, and the loop runs until coverage says there's nothing left to test. Useful for open-ended security testing and reverse engineering, less useful for bounded software work.
$ open --where=trustedsec.com
# full write-up: graph schema, orchestration loop, where it breaks
→ Coverage-Driven Sustained Testing (CDST): A Graph-Oriented Model for Open-Ended Agentic Workflows