SuperAPI AI Directory

tdd

Testing & quality

View repository

Updated 10-15 22:17

About this skill

TDD Persona

Apply the execution contract before this procedure.

Agent Phases

Phase 1: Context & Test Design

  1. load-context: Load schema, routes, and patterns.
  2. plan-tests: Choose the best first failing spec.
  3. write-tests: Write test and verify failure.

HARD GATE — tdd-process (from ruby-core-skills)

  • Test EXISTS and is RUN.
  • FAILS for correct reason (e.g., undefined method 'full_name').
  • If FAIL is incorrect (syntax, config), return to write-tests.

Phase 2: Implementation

  1. Proposal Checkpoint: Propose implementation (e.g., "Concatenate first + last name").
  2. Scope Check: Continue within the existing authorized scope.
  3. Minimal Implement: Smallest change to pass test.
  4. Verify PASS: bundle exec rspec spec/path/to/spec.rb.

If test does not pass, fix minimal changes and re-verify.

Phase 3: Iterate (Optional)

Return to Phase 1 for next behavior or proceed to Phase 4.

Phase 4: Finish

  1. Quality Check: bundle exec rubocop && bundle exec brakeman && bundle exec rspec.
  2. write-yard-docs (from ruby-core-skills): Document public Ruby API.
  3. code-review: Self-review PR diff.
  4. Open PR: Feature complete.

Concrete Example

Abbreviated walkthrough for adding a full_name method to a User model. For the full end-to-end example, see assets/example.md.

Step 1 — Write the failing spec (spec/models/user_spec.rb):

RSpec.describe User, type: :model do
  describe '#full_name' do
    it 'returns first and last name joined by a space' do
      user = User.new(first_name: 'Jane', last_name: 'Doe')
      expect(user.full_name).to eq('Jane Doe')
    end
  end
end

Run: bundle exec rspec spec/models/user_spec.rb Expected failure: NoMethodError: undefined method 'full_name' for #<User ...> ✅

Step 2 — State the approach

Proposal: Add def full_name = "#{first_name} #{last_name}" to app/models/user.rb.

Step 3 — Minimal implementation (app/models/user.rb):

def full_name
  "#{first_name} #{last_name}"
end

Run: bundle exec rspec spec/models/user_spec.rb → 1 example, 0 failures ✅

Step 4 — Quality check:

bundle exec rubocop && bundle exec brakeman && bundle exec rspec

All green → write YARD docs → self-review → open PR.


Output Style

When completing a TDD cycle, produce a report following the template in assets/tdd-report-template.md. At minimum the report must include:

  • RED: spec file path and line, exact failure class and message, confirmation the failure is for the correct reason.
  • Proposal: one-line implementation summary and the authorization scope.
  • GREEN: implementation file path and line range, spec pass confirmation.
  • Iterate: number of additional RED→GREEN cycles and a summary of each.
  • Quality Gate: RuboCop, Brakeman, full RSpec suite, YARD docs, and self-review results.

Integration

Predecessor This Persona Successor
load-context tdd code-review
define-domain-language tdd quality
None (standalone) tdd PR submission

Use plan-tests alone if you only need to decide which test to write next.

Use write-tests alone if the test design is already decided and you only need to implement the spec.


Error Recovery

Test fails for the wrong reason (syntax/config error):

  1. Identify error class — SyntaxError, NameError, LoadError indicate test problems, not missing features.
  2. Fix the test to correctly target the missing behavior and re-run until the failure class is correct (e.g., NoMethodError).

Implementation makes test pass but breaks other tests:

  1. Run bundle exec rspec to identify regressions.
  2. Revise implementation to satisfy both the new test and existing tests.
  3. If impossible, the feature conflicts with existing behavior — discuss with user before proceeding.

Quality gate fails (RuboCop/Brakeman):

  1. Run bundle exec rubocop -a for auto-correctable offenses; fix remaining ones manually.
  2. For Brakeman warnings, assess whether they are false positives — if real, fix before proceeding.