How FALSEWORK works

1. Parse. Each Python file is parsed into a syntax tree. Test files are skipped. A file that does not parse gets only the text checks, and if nothing could be analysed the grade is UNKNOWN, never REAL.
2. Run the checks. Rules over the syntax tree and the comments: where a random number flows, what a function returns on every path, whether a "verify" function compares anything, whether an error handler records success, whether a route uses the tenant it was given, and more. Every check.
3. Assign a severity. Each finding is LOW, MEDIUM, HIGH or CRITICAL. A new check ships at LOW (listed for review, not counted) and is raised only when measurements show it does not raise false alarms.
4. Grade. Any CRITICAL → STUB. Any HIGH or MEDIUM → PARTIAL. Otherwise REAL. Nothing analysed → UNKNOWN.
5. Cross-check routes. When a route file and its engine are scanned together, a route that calls a method the engine does not have is route_engine_mismatch (CRITICAL).

Why no AI model

A model's judgement changes between runs and can be argued with by the code it is reading. A rule cannot. That is what lets KEYSTONE say "a deterministic detector, not the model, decided there was a defect."

What a grade means

FIRME AI CORPORATION · CAIN-42 is a Byzantine trust fabric. Numbers marked "our own check" were run by us, not by an independent party. Nothing on these pages is a certification.