Which checks should I run on a sample before committing to wider use?

Run these checks on a sample before you commit to wider use of a release, and record the result of each so the review can be repeated.

- Every condition links to an existing decision, and every evidence item links to an existing condition. - Timestamps are ordered sensibly: evidence follows the condition opening, and decisions precede their amendments. - Policies do not contradict each other, and each waiver cites a policy that actually permits it. - Each task has an expected action, a forbidden action and a reason code that a practitioner would accept. - Blocked, cleared and ambiguous cases are all present, with no single text cue that cleanly separates them. - Free-text notes agree with the structured state, except where a task deliberately plants a mismatch. - Resetting the environment returns it to the identical starting state on repeated runs of the same task. - Nothing in the content looks lifted from a source you recognise, and any doubt is raised with MirrorFirm.