# Standards rule model

PDF metadata includes SHA-256, page count, text-layer status and import time. Pages without text are marked `OCR_REQUIRED`; no text is invented. Clause extraction preserves clause number, text, page and confidence.

Ollama receives only a small clause context and may return a candidate JSON rule. Candidates remain `New` until human approval. Only `Approved`, `Implemented` or `Tested` rules are evaluated. Profiles contain manifest, approved rules, references and a signature hash.