Shared Foundation Security Considerations

Table of Contents

1. IAM Principles
2. Credential Hygiene
3. RDS Data API Warning
4. Secure Defaults in Examples
5. Output Handling

IAM Principles

The caller’s IAM principal needs read and write permissions for RDS to create and modify clusters. Scope permissions to the minimum required actions.

Required permissions (by service):

Managed policies AmazonRDSReadOnlyAccess and CloudWatchReadOnlyAccess cover most of this; add Pricing and Savings Plans read actions via a scoped custom policy.

Do NOT use AdministratorAccess or *:FullAccess managed policies. Scope write permissions to the specific actions the skill uses: rds:CreateDBCluster, rds:CreateDBInstance, rds:ModifyDBCluster, rds:ModifyDBInstance, rds:AddTagsToResource, rds:RemoveTagsFromResource. For reads: rds:Describe*, rds:List*.

Credential Hygiene

• Prefer short-lived credentials (IAM roles, ada credentials update, SSO) over long-lived IAM user keys.
• Do NOT create or store long-lived DB passwords from within the skill. If the user’s Isengard credentials are expired, prompt them to refresh outside the skill.
• IAM auth tokens are approved. Calling aws rds generate-db-auth-token or rds_client.generate_db_auth_token() is explicitly safe — these produce short-lived (15-minute) tokens derived from the caller’s IAM identity. They are not stored credentials. Use them when IAM database authentication is enabled on the cluster.
• Do NOT log or echo DB passwords or raw secret values. For RDS Data API precheck runs, reference secrets by their secretArn and let the service resolve them.
• For SSM Run Command prechecks, pass DB credentials via inline JSON parameters attached to the Run Command invocation — never via positional filesystem arguments.

RDS Data API Warning

Enabling RDS Data API solely to run upgrade prechecks widens the cluster’s connectivity surface. The Data API endpoint is HTTPS-reachable over the public AWS plane (authenticated with IAM), so it’s safer than opening a new SG ingress rule, but it’s still an additional attack surface.

• Warn the user before recommending they enable Data API for a one-off precheck run
• If the cluster is production and Data API is not already enabled, prefer the user-runs-script precheck method instead
• If Data API is enabled for the workflow, remind the user to disable it after prechecks if it wasn’t previously in use

Secure Defaults in Examples

Any CloudFormation, CDK, or AWS CLI snippet produced by this skill MUST use secure defaults:

• Cluster configuration: StorageEncrypted: true (and KmsKeyId if the user has a customer-managed key)
• TLS: cluster parameter group enables require_secure_transport=ON
• Security groups: scoped CIDR ranges or security-group references — NEVER 0.0.0.0/0 or ::/0
• Public accessibility: NEVER use --publicly-accessible. If the user needs connectivity from outside the VPC, use the RDS Data API (HTTPS + IAM), an EC2 bastion with SSH tunnel, or VPN/Direct Connect into the VPC.
• Parameter groups: do NOT disable general_log, slow_query_log, the audit log (server_audit_logging), or other audit/logging parameters “for convenience”
• Logging & monitoring: recommend enabling CloudTrail so Aurora control-plane API activity (create / modify / delete / failover) is recorded, and CloudWatch alarms on security-relevant metrics such as LoginFailures and DatabaseConnections. CloudWatch log exports (error, slowquery, audit) give query-level visibility but do not cover API-level activity — CloudTrail does.
• Resource names: no prod, production, or PROD as example/default values — those get copy-pasted into production accidentally

Output Handling

• Cost numbers, instance types, and cluster IDs are not sensitive on their own, but combined with account ID they reveal environment topology. When presenting results, don’t unnecessarily include the account ID.
• If a workflow surfaces a secret ARN, show only the ARN, never attempt to resolve it.
• Upgrade precheck findings may include schema names, table names, or query text from the user’s database. If the output is going to be shared (posted to a ticket, shared in chat), warn the user to review for sensitive identifiers before sharing.

References

• Security in Amazon Aurora (opens in a new tab)
• AWS Well-Architected Framework — Security Pillar (opens in a new tab)
• IAM database authentication for Aurora (opens in a new tab)
• Using SSL/TLS with Aurora MySQL (opens in a new tab)