The panel repeatability is evaluated through the product:replicate interaction whereas in the other case (reproducibility), the panel reproducibility is evaluated through the MS error.
For individual panelist performance, the agreement is evaluated through the correlation between the panelist score and the average over the panel. High correlations indicate good agreement. The colour coding helps identify poor judges and/or attributes where the panel struggled
For performance of the whole panel the agreement is evaluated through the p-value of the product by panellist interaction.
You can also find information regarding interpretation of panel performance in EyeOpenR user guide (https://sites.google.com/a/eyequestion.nl/eyeopener/analysis/panel-performance/panelist-performance/c-interpretation)