Detection has limits
Detection uses field knowledge (Secret data, env var names, the last-applied annotation) and token patterns (private keys, cloud access keys, JWTs, credentials embedded in URLs). It will miss secrets with unusual shapes and flag some harmless values, so review the preview before sharing.
Redaction rewrites only the selected scalar values and keeps comments and structure elsewhere. The original stays in the editor, and the redaction is a single step you can undo.
Questions
Why is last-applied-configuration flagged?
kubectl apply stores the full applied object in that annotation. For a Secret, that includes its data, so redacting the data field alone is not enough.
Is the redacted export safe to post publicly?
Only after you review it. Names, hostnames, image registries and namespaces can be sensitive too, and they are not redacted automatically.