The DSG review process has established a precise spoiler prevention standard: no readable daily count, no yellow prompt text, no option text, no final order, and no post-solve information visible in any frame. GPT-5.5's rejection criteria have become more granular with each review cycle — moving from general spoiler concern (v1) to specific mask leakage of count-feedback and prompt text (v2)