Nunjucks Best Practices and Tips

Nunjucks is a core principal when creating and updating Calculated policies. Basic nunjuck template documentation can be found over on Mozilla's github page.

White Spaces

White space management is critical when building good nunjuck templates. If your calculated policy is not behaving as expected, be sure to review the template for any extra or missing white spaces.

  • Do use white spaces consistently.

  • Create balance by allowing white spaces before a block, but not after it. For example:

    • {% ... -%} ... {% ... -%}
  • Do not leave spaces at the end of lines.

  • Do not create blank lines with only spaces in them.

  • Do not accidentally combine lines together by removing white spaces.


Turbot Guardrails GraphQL responses are either individual key value pairs or an array of dictionaries that contain keys and values. In the latter case, an incorrect attempt to output the array will result in [object Object] (and will repeat for the amount of objects in the array, if applicable).

In order to display actual values, it is necessary to go one step further. For this example, check out this query:

If we request the output of the tags using the method in the first screenshot, we get the multi [object Object] response. Instead, what can be done is requesting the specific index of the tag in the array. Keep in mind that with a graphql response like above, each key value tag pair is its own dictionary object. This means to get a specific tag key or value, we need to call the corresponding index in the query result:

If we want to output all tags associated with the resource, we can easily use a for loop to pull out all values:

This concept applies to any attribute stored as an array. Further, this for loop can be used to do things like check a resource for the existence of a specific tag, and if it does not exist, the resource is deemed Not approved. In this example we are checking for the existence of a tag with the key test:

A tag with the key test does in fact exist, so this AMI is approved!

Alternatively, one can use the following query to get a different type of response:

Using this method, it is trivial to output the tag value:

This does NOT work if the tag has a hyphen, however. Attempting to change the template from {{ $.resource.turbot.tags.test }} to {{ $.resource.turbot.tags.test-tag }} results in a NaN value. Instead, we can use the syntax {{ $.resource.turbot.tags['test-tag'] }}.

Note that the turbot guardrails query results in a dictionary, so the {{ $.resource.turbot.tags['key'] }} syntax can be used for values that do not have a hyphen in the key.


Guardrails calculated policies support the use of RegEx to create filters within the template logic.

This enables administrators to do things like requiring a specific tag to be in email format or check the names of resources based off specific patterns.

For example, let's assume that we want to ensure that the tag Owner has a value that is in email format, or, where must exist as well as any number of characters prior to the @ symbol.

First, query for the tags.

Using the template, add a Regex statement. Refer to one of the many regex testers found online for more information, such as In this case, the statement says any string may occur prior to the @ symbol - upper and lower case letters as well as numbers, and the characters ., _, -, and +. The string must follow the initial string.

If valid, the template will simply return the tag key and existing value. If it is not valid, the template outputs a new tag asking the user to add a valid email.

Common Errors

The Calculated Policy Builder gives real time feedback into the query and template entered. While it is often recommended to create policies via Terraform, the builder can be invaluable in testing policies to ensure they do what you think they do.

Template Output Errors

This occurs when the template is missing a closing or opening bracket for a variable. Variables in Nunjucks must be passed with the format {{ $variableName }}.

The template has an if statement but it has not been properly closed. Refer to the official Nunjucks documentation for more information on if statements.

Very similar to the previous error, this occurs when a for statement has not been properly closed. Refer to the official Nunjucks documentation for more information on for statements.

These are just a few of the most common errors when creating calculated policies.

Query Output Errors

While not directly Nunjucks, knowing what GraphQL errors mean can significantly reduce trouble shooting time. The GraphQL query window is a bit more friendly with respect to errors, as extra or "rogue" brackets will be highlighted with the red underline. This manifests itself in two different error messages:

Review your query to ensure all opening brackets have an associated closing bracket.

Queries do have a defined schema. If that schema is broken in the GraphQL input, the following error will be displayed.

This error is more specific and will often offer guidance on what to try next. Check out our 7 minute GraphQL lab for a primer on GraphQL queries.