Simulated Reader Review

Generate editorial hypotheses from explicit audience lenses. A model persona is not a human participant; agreement, emotion, predicted navigation, and synthetic acceptance do not establish real audience response or lived experience.

Select the smallest useful experiment

Infer the authorized scope and budget from the request. For an ordinary review, use a small set of distinct lenses in one pass. Use sequential cold-reading only when order of information, expectation mismatch, or navigation is the uncertainty. Use extra agents, collision exercises, or large panels only when their additional coverage justifies the cost. Ask about budget only when it is consequential and not already agreed; do not repeat authorization at routine phases.

Read relevant sections of process.md, not the entire process in every role. config-template.md defines the config; config.md is an illustrative larger panel, not the default.