Developer Builds AST-Based Scorer to Grade Python Retry Loop Safety

A developer created a static code analysis tool called a 'hygiene scorer' that reads Python source files and grades retry loop implementations for safety properties. The scorer walks the abstract syntax tree (AST) and awards points for features like attempt caps, wall-clock timeouts, backoff, jitter, and idempotency keys, while penalizing unbounded loops. The motivation was to address a common pattern in agent and automation code where retry loops lack hard stop conditions, potentially causing runaway processes. The tool was built and tested using a free AI model accessed via MonkeyCode, with outputs validated against fixture files rather than trusted outright. The author argues that verifying whether a loop can terminate is a more practical engineering concern than evaluating an AI model's reasoning quality.