Criterion measurements for selected prompt filtering, PII masking, and token-estimation functions. Results are commit- and runner-specific.
Measures OpenAI-compatible schema generation for an explicitly registered and policy-allowed local tool.
AiGuardrails::mask_pii() applies bounded heuristics to supported identifier patterns; it is not complete DLP.
A heuristic prompt-size estimate; it is not a provider tokenizer or billing authority.