1. Manually programming this to automatically get meaningful feedback directly related to the documentation would require an agent or sifting through error logs. Otherwise, coding the patterns equivalent to support documentation is tedious manually, perhaps does not outweigh the benefits.

2. Without the ability to update the documentation directly, the system yields little benefit. Perhaps and agent trained on the codebase can (more efficiently) use automation tests based on the flow pattern rather than a visual browser agent, but it would still need the agent architecture.