Hi ​@swapnil, I have two suggestions for you to consider. Let me know if you’ve already tried these and I can maybe suggest other things to consider:

• You can use the cloud filter tool in the Subscriptions API. This filters imagery matching your criteria based on an area of interest cloud percentage statistic, as opposed to scene level (larger area) metadata. Note: If you are using the Third Party Data Import API from Sentinel Hub, you can switch to the Subscriptions API for this feature.
• You can deliver data with the Usable Data Mask along with the PlanetScope imagery to your collections. This can allow for you to access the cloud predictions in your evalscript to filter out unusable data.  

To answer your questions:

• Does Planet imagery delivered through Sentinel Hub include unfiltered/cloudy scenes by default, or is there internal cloud filtering applied in AUM subscriptions?

  • It does by default, but you can apply the filtering shared above.

• What are the best practices or parameters we can use (in Sentinel Hub or Planet APIs) to ensure we only receive cloud-free or low-cloud scenes (e.g., <10%)?

  • You can specify your desired percentage in the Subscriptions API shared above. You can also mask out pixels in the Processing or Statistical API by accessing those Usable Data Mask bands.

• Is there any difference in cloud-filter behavior between PlanetScope and SkySat products in this context?

  • Functionally, the cloud filtering work the same. For both PlanetScope and SkySat we calculate the Usable Data Mask and use it to derive scene level metadata for both data products that can be used to search in the API.