ActionFigure actions return plain hashes, making them straightforward to test without controller setup or request scaffolding. You call the action directly, receive a result, and assert against it.

Both Minitest and RSpec helpers are provided. They wrap status checks in expressive, intention-revealing assertions so your tests read clearly.

Require the helper and include the module in your test class:

require "action_figure/testing/minitest"

class Users::CreateActionTest < Minitest::Test
  include ActionFigure::Testing::Minitest
end

Each status assertion has a negated counterpart — refute_Ok, refute_Created, … — that passes when the status is anything other than the named one. Statuses added with ActionFigure.register_error get matching assert_*/refute_* helpers automatically, whether registered before or after this adapter loads.

These helpers compare only result[:status] against the Rack-style symbol Rails uses in render — they do not assert on [:json] keys, payloads, or error message text. Combine them with assertions on result[:json] (or matchers on the body your formatter produces) whenever shape matters.

All assertions accept an optional second argument for a custom failure message:

assert_Ok(result, "expected the user to be created successfully")

When a status assertion fails, the default message shows the expected and actual status:

Expected result status to be :ok, but got :unprocessable_content

Use assert_action_json to match a (possibly nested) subset of result[:json] — the Minitest counterpart to RSpec's have_action_json. Nested Hashes match as subsets, and Regexp values match against strings:

assert_action_json(result, status: "success")
assert_action_json(result, status: "success", data: { name: "Jane" })
assert_action_json(result, data: { email: /@example\.com\z/ })

refute_action_json passes when the fragment does not match. Both fail with a clear message when given a non-result value or a hash missing the :json key.

Require the helper in your spec support file. No include is needed -- the matchers are registered globally:

# spec/spec_helper.rb
require "action_figure/testing/rspec"

Load order: require this library after RSpec Core and expectations load (usual practice: append it toward the bottom of spec/spec_helper.rb, after any require "rails_helper" / RSpec.configure boilerplate from your app). ActionFigure pulls in rspec/matchers; minimalist scripts without the full rspec CLI shim must require "rspec/expectations" (and typically require "rspec/core") before this file.

Like the Minitest helpers, each be_* matcher compares only result[:status] — [:json] is ignored unless you assert on it separately. Use have_action_json when you want a focused assertion against the json body (compose with a_hash_including for nested subsets):

expect(result).to be_Ok
expect(result).to have_action_json(status: "success")
expect(result).to have_action_json(
  status: "success",
  data: a_hash_including(name: "Jane")
)

Matchers support negation:

expect(result).to be_Ok
expect(result).not_to be_Forbidden

Failure messages mirror the Minitest style:

expected result status to be :ok, but got :unprocessable_content

The examples below use Minitest, but the same patterns apply to RSpec with the corresponding matchers.

The examples below use the JSend formatter (ActionFigure[:jsend]) for consistency. The structure of result[:json] depends on your chosen formatter — see Response Formatters for the shape each format produces.

Call your class and assert both the status and the returned data:

class Users::CreateActionTest < Minitest::Test
  include ActionFigure::Testing::Minitest

  def test_creates_a_user
    result = Users::CreateAction.create(params: { email: "jane@example.com", name: "Jane" })

    assert_Ok(result)
    assert_equal "jane@example.com", result[:json][:data][:email]
    assert_equal "Jane", result[:json][:data][:name]
  end
end

When testing validation failures, assert both the status and the error message content. Testing only the status is insufficient -- it does not prove the correct validation failed.

class Users::CreateActionTest < Minitest::Test
  include ActionFigure::Testing::Minitest

  def test_rejects_missing_email
    result = Users::CreateAction.create(params: { name: "Jane" })

    assert_UnprocessableContent(result)
    assert_includes result[:json][:data][:email], "is missing"
  end
end

Actions often receive context such as current_user: as keyword arguments alongside params:. Pass them directly in the test:

class Posts::CreateActionTest < Minitest::Test
  include ActionFigure::Testing::Minitest

  def test_creates_a_post_for_the_current_user
    user = users(:jane)
    result = Posts::CreateAction.create(params: { title: "Hello", body: "World" }, current_user: user)

    assert_Created(result)
    assert_equal user.id, result[:json][:data][:author_id]
  end
end

Call the action using its discovered method name:

class Products::SearchActionTest < Minitest::Test
  include ActionFigure::Testing::Minitest

  # class SearchAction
  #   include ActionFigure[:jsend]
  #
  #   params_schema do
  #     required(:query).filled(:string)
  #   end
  #
  #   def search(params:, **)
  #     products = Product.where("name ILIKE ?", "%#{params[:query]}%")
  #     Ok(resource: products)
  #   end
  # end
  def test_finds_matching_products
    result = SearchAction.search(params: { query: "keyboard" })

    assert_Ok(result)
    assert result[:json][:data].any?, "expected at least one matching product"
  end
end

Actions that perform side effects without returning data use NoContent():

class Sessions::DestroyActionTest < Minitest::Test
  include ActionFigure::Testing::Minitest

  # class Sessions::DestroyAction
  #   include ActionFigure[:jsend]
  #
  #   def destroy(session:)
  #     session.destroy!
  #     NoContent()
  #   end
  # end
  def test_destroys_the_session
    session = sessions(:active)
    result = Sessions::DestroyAction.destroy(session: session)

    assert_NoContent(result)
  end
end

Every action that defines a params_schema exposes the underlying validation contract via .contract. This returns a Dry::Validation::Contract instance that you can call directly -- useful for validating input without executing the action.

contract = Users::CreateAction.contract
result = contract.call(email: "jane@example.com", name: "Jane")

result.success?    # => true
result.to_h        # => { email: "jane@example.com", name: "Jane" }

When validation fails, inspect the errors:

result = Users::CreateAction.contract.call(email: "", name: "Jane")

result.failure?      # => true
result.errors.to_h   # => { email: ["must be filled"] }

This runs both the schema and any rules defined on the action -- the same validation pipeline that the class-level trigger uses, without the side effects.

Actions that do not define a params_schema return nil from .contract.

The contract exposes the schema and rules for introspection:

contract = Users::CreateAction.contract

contract.schema                        # => the Dry::Schema::Params instance
contract.schema.key_map.map(&:name)    # => ["email", "name"]

contract.rules                         # => array of Dry::Validation::Rule objects
contract.rules.map(&:keys)            # => [[:email]]

This is useful for building documentation generators, admin panels, or debugging which validations an action enforces.

• Form validation endpoints -- validate input and return errors without creating or modifying resources.
• Testing validation rules in isolation -- assert that specific inputs produce specific errors without needing to stub dependencies that #call would use.
• REPL exploration -- inspect what an action expects by calling its contract interactively.

class Users::CreateActionTest < Minitest::Test
  def test_email_is_required
    result = Users::CreateAction.contract.call(name: "Jane")

    assert result.failure?
    assert_includes result.errors.to_h[:email], "is missing"
  end
end

The testing adapters wrap the .contract.call boilerplate above in intention-revealing helpers. They are formatter-agnostic — they exercise the validation pipeline directly, so the same assertions work regardless of which formatter the action includes.

Minitest — the subject is the action class:

class Users::CreateActionTest < Minitest::Test
  include ActionFigure::Testing::Minitest

  def test_accepts_valid_params
    assert_valid_params(Users::CreateAction, { email: "jane@example.com", name: "Jane" })
  end

  def test_requires_email
    # passes when the contract rejects the params at all
    assert_invalid_params(Users::CreateAction, { name: "Jane" })

    # scope to a field: passes only when :email is among the errors
    assert_invalid_params(Users::CreateAction, { name: "Jane" }, on: :email)
  end
end

RSpec — the subject is the action class, not a result hash:

RSpec.describe Users::CreateAction do
  it "accepts valid params" do
    expect(Users::CreateAction).to accept_params(email: "jane@example.com", name: "Jane")
  end

  it "requires email" do
    expect(Users::CreateAction).to reject_params(name: "Jane")
    expect(Users::CreateAction).to reject_params(name: "Jane").with_error_on(:email)
  end
end

Both adapters raise a clear ArgumentError when the action class declares no params_schema (and therefore has no contract to validate against).

Validation errors vs. error bodies. These helpers are the right tool for asserting validation behavior. There is no formatter-agnostic helper for non-validation error bodies (e.g. a NotFound/Conflict you return with a custom errors: payload) — each formatter stores those differently and the result hash carries no formatter identity. Assert those with a format-specific assert_action_json / have_action_json.

• Assert fully -- for validation and rule failures, assert both the HTTP status and the error message. Testing only the status does not prove the correct validation failed.
• Named locals, not subject -- use a descriptive local variable (result, action) in each test instead of a shared subject helper method.
• Use return values for assertions -- assert on what Ok(resource: ...) returns rather than capturing outer variables with closures.