Model Test

Write or review the spec for the model at $ARGUMENTS. First read .claude/skills/rspec-test-expert/test-quality.md: the fat catalogue, the proof (coverage plus probes), concurrency-safe spec runs, and the review workflow all live there. This file covers what is particular to models.

Map paths: app/models/foo/bar.rb ↔ spec/models/foo/bar_spec.rb; app/models/concerns/x.rb ↔ spec/models/concerns/x_spec.rb. If the spec exists, you are in review mode (the workflow in test-quality.md); otherwise write mode.

What a model's behavior is

Read the model and list these. They are what the spec must guard:

Model-specific fat to remove

• One-liners for every association and every presence validation, copied from the class body.
• "Factory creates a valid record", trait loops, build vs create demos.
• Re-testing concern behavior (HipaaLoggable, SoftDeletable, Normalizes…) in every includer. Test the concern once in spec/models/concerns/; in the includer, test only its configuration.
• respond_to(:soft_delete), included_modules, column_names checks.
• N+1 or "query count" tests on plain associations. Keep a query-count example only when the model has a method whose point is preloading.
• create-heavy setup where build validates fine. Only scopes, callbacks, and uniqueness need persisted rows.

Shape

# frozen_string_literal: true

require "rails_helper"

RSpec.describe InvoiceAdjustment, type: :model do
  describe "validations" do
    it { is_expected.to validate_numericality_of(:amount_cents).other_than(0) }

    it "rejects a write-off larger than the invoice balance" do
      invoice = create(:invoice, balance_cents: 5_000)
      adjustment = build(:invoice_adjustment, invoice:, kind: :write_off, amount_cents: -6_000)

      expect(adjustment).not_to be_valid
      expect(adjustment.errors[:amount_cents]).to include("cannot exceed the remaining balance")
    end
  end

  describe ".posted_between" do
    it "includes both boundary days and excludes the day after", :aggregate_failures do
      travel_to Time.zone.local(2026, 9, 1, 12) do
        first = create(:invoice_adjustment, posted_on: Date.new(2026, 8, 1))
        last = create(:invoice_adjustment, posted_on: Date.new(2026, 8, 31))
        after = create(:invoice_adjustment, posted_on: Date.new(2026, 9, 1))

        result = described_class.posted_between(Date.new(2026, 8, 1), Date.new(2026, 8, 31))
        expect(result).to contain_exactly(first, last)
        expect(result).not_to include(after)
      end
    end
  end
end

(Illustrative: check the real model for names, messages, and columns. Never copy expectations from this example.)

Prove it, then report

Follow test-quality.md: the spec is green, coverage --baseline shows no lost_lines, and 2–5 probes are aimed at custom validations, scope conditions, callback effects, and transition guards, all KILLED. Then lint with bin/standardrb <spec> and report as test-quality.md describes (or in the caller's format).