A Blueprint apply replaces the target account's guardrails, without any notice in the apply dialog. In our test account we had set two guardrails with their own replacement texts; after "Apply latest version" only the source account's guardrail was left. Words an account has blocked for itself are lost on every apply, and a word blocked for one account in the source account reaches every other account.

Request: guardrails as their own component in the Blueprint apply dialog. Ideally: fixed words from the Blueprint plus protected words per account, kept on every apply.

Workaround today: we keep only words that apply to every account in the source account and set each account's own words again after every apply.