What is the correct way to re-use boostrap deployment roles?

[rupe120]

I am attempting to create a CodePipeline for a fork of the AWS Solution Innovation Sandox on AWS app https://github.com/aws-solutions/innovation-sandbox-on-aws

I am trying to modify that solution as little as possible, while also minimizing the number of resources I create. I created a second CDK repository that builds a CodePipeline to run the npm run deploy:xxx commands. In the Buildspec that I added to the Innovation Sandbox repo I assume the CDK deploy role before each of the 4 deploy run statements.

The problem is when the deployment runs it get the following errors

current credentials could not be used to assume 'arn:aws:iam::96xxx57:role/cdk-hnb659fds-file-publishing-role-96xxx57-us-east-1', but are for the right account. Proceeding anyway.

InnovationSandbox-AccountPool: fail: User: arn:aws:sts::96xxx57:assumed-role/cdk-hnb659fds-deploy-role-96xxx57-us-east-1/CodeBuild-Deploy-AccountPool-1769178658 is not authorized to perform: s3:PutObject on resource: "arn:aws:s3:::cdk-hnb659fds-assets-96xxx57-us-east-1/35658981a4fc9561802559f1251f57680065e5f463cc1cd95d11ca6654216d59.json" because no identity-based policy allows the s3:PutObject action

It seems like I actually need to assume both the File Publishing and the Deployment roles, but you can only assume one role at a time.
Should I just create my own roles to assume? I was just also trying to not create more resources than I need and re-use what CDK bootstrap had already setup.