Cannot use source to execute Bash script on zsh - bash

Question

I have a Mac, on which I have installed and configured zsh to be my default shell. It has worked very well with simple Bash scripts. Until today.
I was trying to set up my AWS CLI access with MFA and used a script to do the same. Since I have multiple accounts, I used an aws_accounts file to store the account numbers with 400 permissions on it. (No real reason why, just felt like it and it works).
I then source this file and get to the part where I need to provide my MFA key, and those work too. The last step, where I export the access key, secret key and session token, works as long as the script is executing, but once done, echoing it shows a blank, because it now no longer has set it in the current shell.
I know the workaround is to do source aws_mfa_access default to run the script, but it doesn't work and I get a bad substitution error. I've tried various combinations of #!/usr/bin/env bash, #!/bin/bash and #!/usr/bin/env zsh, but to no avail. What's going on?
[aws_account_numbers]
default_account_number=<number>
sandbox_account_number=<number>
production_account_number=<number>
[aws_mfa_access]
#!/usr/bin/env bash
source aws_account_numbers
AWS_ENV="${1}"
AWS_ACCT_NBR="${AWS_ENV}"_account_number
#export your aws access key that matches with your account, username
export AWS_PROFILE="${AWS_ENV}"
#If there are existing environment variables set, this can cause issues so we unset them first
unset AWS_SESSION_TOKEN
unset AWS_SECRET_ACCESS_KEY
unset AWS_ACCESS_KEY_ID
#Set the serial number of your MFA token
MFA="arn:aws:iam::${!AWS_ACCT_NBR}:mfa/codingincircles"
#Get the code from the MFA device
echo "Please enter the MFA code for the ${AWS_ENV} account: "
read -r code
#Get the credentials from AWS and store the response in a variable
creds="$( aws sts get-session-token --serial-number "${MFA}" --token-code "${code}" )"
#Parse the response into separate variables
AWS_ACCESS_KEY_ID="$( echo "${creds}" | jq -r .Credentials.AccessKeyId )"
AWS_SECRET_ACCESS_KEY="$( echo "${creds}" | jq -r .Credentials.SecretAccessKey )"
AWS_SESSION_TOKEN="$( echo "${creds}" | jq -r .Credentials.SessionToken )"
#Display the keys to the user for reference/confirm proper working
echo "${access_key}"
echo "${secret_key}"
echo "${session_token}"
#Set the appropriate environment variables
export AWS_ACCESS_KEY_ID="${AWS_ACCESS_KEY_ID}"
export AWS_SECRET_ACCESS_KEY="${AWS_SECRET_ACCESS_KEY}"
export AWS_SESSION_TOKEN="${AWS_SESSION_TOKEN}"
EDIT: Two things of note:
The parameter substitution is not an issue at all. It works just fine as is, though the other methods suggested in the comments and answer work too. (I used the jq -r tip and it works like a charm! Thank you!)
I removed the source command in my script (on line 3) and then was able to, without any errors of any kind, able to invoke my script as source aws_mfa_access default. The exported variables persisted and I am able to use the CLI with no problems.
So why does this not like me using source in my script? I've also edited the script to reflect some of the changes.

Answer 1

Indirect parameter expansion is different in zsh than it is in bash
MFA="arn:aws:iam::${(P)AWS_ACCT_NBR}:mfa/codingincircles"
However, you can avoid indirect parameter expansion in the first place by using an associative array to store the account numbers.
typeset -A account_numbers
account_numbers[default]=...
account_numbers[sandbox]=...
account_numbers[production]=...
Then
MFA="arn:aws:iam::${account_numbers[$1]}:mfa/codingincircles"
Finally, source is not an external command that means "execute a bash script here". It's a shell built-in that executes a file in the current shell. If the current shell is bash, it will attempt to execute the contents of a file as a bash script; if it's zsh, as a zsh script; if it's dash, as a dash script.