Review security in the mlschool CloudFormation files

[svpino]

Make sure we assign the least amount of privileges possible to the role created in this file.

[traycerai]

Plan

Observations

The IAM role defined in cloud-formation/mlschool-cfn.yaml currently uses broad “Action” lists with Resource “*” for many services (CloudFormation, Batch, Events, IAM, EC2, ECR, CloudWatch Logs, SageMaker, S3, States). Our codebase only directly calls a subset (SSM, S3, SageMaker via MLflow/Sagemaker backend, STS AssumeRole).

Approach

1. Inventory actual AWS calls in your Python and backend code. 2. Map each required action to precise resource ARNs or patterns. 3. Edit the inline policy in mlschool-cfn.yaml: remove unused service blocks, tighten Resources, add Conditions for PassRole. 4. Validate templates with cfn-lint/cfn-nag, deploy to a dev account, run pipelines, watch for AccessDenied errors.

Reasoning

I parsed your CloudFormation templates to identify the IAM role and its policies, then scanned key pipeline and backend code to determine which AWS services and actions are truly used (SSM:GetParameter, S3 operations, SageMaker model/endpoint actions, STS:AssumeRole).

Proposed File Changes

📄 cloud-formation/mlschool-cfn.yaml ^(MODIFY) 🔗

Referred Files

• 📄 cloud-formation/mlschool-cfn.yaml ^(MODIFY)

• Remove entire statements for services not used by code (Batch, Events, EC2, ECR, CloudWatch Logs, CloudFormation, States).
• Under the IAM statement block, change Resource: "*" to the role’s own ARN via !GetAtt Role.Arn and add a Condition limiting iam:PassedToService to sagemaker.amazonaws.com.
• Add a new SSM block that allows ssm:GetParameter/ssm:GetParameters on exactly:

  - arn:aws:ssm:${AWS::Region}:${AWS::AccountId}:parameter/ec2/keypair/*
  - arn:aws:ssm:${AWS::Region}:${AWS::AccountId}:parameter/aws/service/canonical/ubuntu/server/*

• Add an S3 block allowing only:

  - arn:aws:s3:::mlschool-${AWS::AccountId}
  - arn:aws:s3:::mlschool-${AWS::AccountId}/*

• Add a SageMaker block permitting only the actions your backend uses (ListEndpoints, DescribeEndpoint, CreateEndpoint, UpdateEndpoint, DeleteEndpoint, CreateModel, CreateEndpointConfig, DescribeModel, ListTags, AddTags, InvokeEndpoint) on:

  - arn:aws:sagemaker:${AWS::Region}:${AWS::AccountId}:endpoint/*
  - arn:aws:sagemaker:${AWS::Region}:${AWS::AccountId}:model/*
  - arn:aws:sagemaker:${AWS::Region}:${AWS::AccountId}:endpoint-config/*

Import In IDE

Developer Humor

“Why did the IAM policy go to therapy?
It had too many trust issues! 😂
But now, after least-privilege refactoring,
it’s feeling right at home.”

---

Execution Information

Branch: ai
Commit: c1c3cdb

---

💡 Tips

Supported Commands (Inside Comments)

• Use @traycerai generate to iterate on the previous version of the implementation plan.

Supported Commands (Inside Description)

• Add @traycerai ignore anywhere in the ticket description to prevent this ticket from being processed.
• Add @traycerai branch:<branch-name> anywhere in the ticket description to specify the target branch for the implementation plan.

Community

• Join our Discord Community to get help, request features, and share feedback.
• Follow us on X/Twitter for updates and announcements.