I had the framework audited: every finding was the same shapeEvery finding in a security audit reduced to untrusted input reaching a powerful operation unchecked. One checked chokepoint per boundary fixes it.
