Diagnosing a failure in a thirty-year-old program has usually meant finding the one engineer who can read the storage by eye. These diagnostics put the same evidence in front of anyone on the team, and in front of an AI agent, in the terms the program itself uses. The agent reasons from what the runtime recorded, and its conclusion can be checked against the same record.
Stated limits. These are diagnostics for analysis. They are not a development environment and they do not replace one. They depend on Rocket Visual COBOL: the compiler output and the runtime evidence they read are Rocket’s.