Automated controls require a large number of configuration settings from the biggest decisions (e.g. allowed services) right down to small, but critical, details (e.g. automated tags for cost management). In Guardrails, Policies are used to manage these settings.

In Guardrails, policies provide:

  • Clearly defined, validated and managed definitions for all Resources managed by Guardrails
  • Secure enforcement of required settings across large environments.
  • Exception management, including automatic expiration
  • Shared defaults via recommended settings

Guardrails Policies can be managed using the Guardrails UI, Guardrails API or software configuration management tools.

The v5 mods directory is an invaluable resource for looking up policy URIs, policy values and policy defaults. (A free, self-registered account is required.)

A control objective is a business need, for example at rest enforcement for S3, EC2, and RDS. Note that a control objective may require one or many policies to fully implement.

Key Concepts

A Policy Type targets one of more resource types. This defines which type of resources the setting applies to.

The GCP > Storage > Bucket > Approved policy applies to GCP > Storage > Bucket resource types.

While the policy type targets a specific resource type, you may set the policy at any scope in the Policy Hierarchy at or above the resource. Organizations can define policies such as restricted regions for a folder in Guardrails which are then inherited by accounts and resources within said folder.

Setting the GCP > Storage > Bucket > Approved policy at the Project scope will impact all of the buckets in that Project.

Policy settings have a precedence, which defines whether they are required or recommended on descendent resources.

The Policy Setting is the desired value for the policy. For example, an organization could define specific approved RDS database engine types, which are three simple policies. Alternatively, another organization might want to tag Azure resources with Terraform, where the power of policies are flexed, utilizing Terraform deployment, Guardrails Files, and Calculated policies.

Every policy can have an expiration, after which the policy will no longer be in effect.

Controls implement policies. The policy setting is created, an applicable resource inherits the value, and the relevant control will check and/ or remediate the resource.


Consider an S3 bucket Resource called my-bucket, an instance of the Resource Type AWS > S3 > Bucket. To define the correct configuration of my-bucket, a few policy settings are required. For example:

TypeSetting to define for my-bucket
AWS > S3 > Bucket > Approved"Enforce: Delete if new & empty"
AWS > S3 > Bucket > Approved > Regions[ "us-*" ]
AWS > S3 > Bucket > Encryption at Rest"Enforce: AWS SSE or higher"

The above policy settings are then applied directly to the bucket (if the policies were set at the bucket resource level) as a value or are inherited as a value (the typical situation). In this case, we are telling Guardrails to Enforce: Delete if new & empty if either encryption is not configured to be AWS: SSE or higher OR the bucket is created in a region outside of the US.