Kulinaria AR Re-Analysis: Dataset and Reproducibility Package

This deposit contains no new measurements. It holds nine binary observations transcribed verbatim from the functional test table published in Mardhiyyah et al. (2023), ICITCOM, DOI 10.1109/ICITCOM60176.2023.10442988, together with analysis code that derives exact binomial confidence bounds, design-space coverage, and a geometric operating envelope from them. The accompanying manuscript is an analytical re-examination: it treats the nine-observation table as the object of study and quantifies what it does and does not establish. Key derived results are that nine consecutive successes bound the true success probability only at 66.4 percent (95 percent Clopper-Pearson lower limit), admit a failure rate as high as 28.3 percent, and that 59 zero-failure trials would be required to support a 95 percent reliability claim. A one-factor-at-a-time protocol covers at most 9 of 27 joint factor cells (33.3 percent) and leaves all two-factor interactions inestimable. A pinhole-projection model shows the reported 60-90 degree acceptance band to be geometrically degenerate at its upper edge, where projected marker area reaches zero. Users should not cite this deposit as evidence of the system's reliability. Its central finding is the opposite. Known limitations of the primary data, including unknown replication counts and unreported co-factor levels, are documented in README.md. Code is released under MIT; data, figures and documentation under CC BY 4.0.