Directive Reference

Alphabetical list of TOML directives used in models/, compliance/, output/, and actions/.

`[[create_resource]]`

Creates a new resource based on the origin_resource or header lists.

Directive	Description
`resource_type`	The type of the new resource.
`relation_type`	The type of the relationship from the origin resource.
`name`	String template for the primary key.
`properties`	Key-value table of properties.
`relation_properties`	Key-value table for edge properties.
`match_on`	Array of operators to conditionally apply rule.
`create_from`	Creates via `{ property = "..." }` or `{ list = "..." }`.
`copy_from`	Pulls properties from a relation dynamically.

Links existing resources (the “pull” join).

Directive	Description
`with`	The type of the remote resource.
`join`	`{ local = "...", remote = "..." }` join condition.
`match_with`	Array of operators to filter the remote resources.
`copy_properties`	Array of properties to copy from the target.
`create_relation`	`{ type = "..." }` relation to create.

Pushes data to a connected resource.

Directive	Description
`to`	The destination resource type.
`properties`	Array of strings or tables `{ from, as, template }`.

Overrides asset implicit relation creation.

Directive	Description
`property_key`	Column header name.
`new_type`	The overridden relationship type.
`target_resource_type`	The overridden target resource type.

Defines a compliance rule in compliance/.

Directive	Description
`id`	Standard identifier.
`name`	Human-readable name.
`[[control.target]]`	Array of mutation targets (`origin_resource_types`).
`[control.config]`	Table of values applied as compliance payload.

Generates an artifact in output/.

Directive	Description
`resource_type`	Output resource type classification.
`name`	Primary key of the generated artifact node.
`filename`	Downloadable filename.
`mimetype`	Content MIME type.
`template`	Tera template string rendering JSON.
`jsonnet`	Jsonnet evaluation string rendering JSON.

Defines an executable business logic action.

Directive	Description
`name`	The name of the action.
`description`	Human-readable description, exposed to MCP Agents.
`[[input]]`	Defines the expected input schema (name, description, type).
`[runtime]`	Execution environment configuration (`engine`, `flake`, `files`).
`[[mount]]`	Mounts generated output artifacts from the graph context.
`[env]`	Secure credentials mapping (fetches from Vault).
`exec`	Command array to execute the action.