Every scanner finds a public bucket. That is good, and it is also the easy part. The stores that cause real trouble are private, and readable by forty roles nobody can explain.
Private only means the internet cannot read it directly. It says nothing about the roles, build jobs and old integrations that can.
Sort by what is inside
We read names, tags and structure, and we leave file contents where they are. That is enough to sort most stores into three groups.
- Customer records and anything regulated.
- Backups, which are copies of the first group.
- Logs and build output, which often hold secrets.
Backups are production data with less supervision.
Then count the readers
For each store in the first two groups, count who can read it and how many hops away they are. The number is usually a surprise, and it gives you a clear target for the next sprint.




