The artifact a diagnostic engagement produces. Contents specification:
- Observation log — what evaluators said and recommended, verbatim, instrument-stamped (evaluator, version, access mode, date)
- Sampling conditions — N, prompt variants, repeats, variance; printed in the report body, not an appendix
- Requirement coverage read — which category requirements your representations substantiate, fail to substantiate, or leave invisible
- Confidence breakdown map — where evaluator narration about you hedges, contradicts itself, or requests evidence
- Declared limits — what this report cannot tell you (mechanism; persistence across model updates)
What it does not contain: rankings promises, optimization tricks, competitor disparagement, or any number we haven't earned the digits of.
Example status: none yet. When a real (consented, redacted) example exists it will be linked here; an illustrative mock would be rendered in the SPECIMEN register, never in the visual language of real findings.