Build

When to use

Use for executing focused implementation tasks from an established plan. Not for broad architectural design, planning, or large-scale refactoring.

Process

1. Read: Review the pre-defined task, spec, and relevant code. Stop and clarify if the scope is vague or exceeds a single task.
2. Implement: Apply the smallest useful code change that satisfies the task requirements.
3. Verify:
   • TDD: Write/update tests for new behavior or bug fixes.
   • Reproduction: For bugs, confirm the failure with a test before fixing.
   • Full Suite: Run the strongest available project checks (e.g. bundle exec rspec) to prevent regressions.
   • Manual Check: Perform task-specific verification (e.g., curl an endpoint or check logs).
4. Final Gate: Confirm behavior matches the task, no unrelated scope was added, and all tests pass.

Example: TDD Loop (Code-Level)

Task: "Implement User#admin? based on the role attribute."

1. Write failing spec:

   # spec/models/user_spec.rb
   it "returns true if the user is an admin" do
     user = User.new(role: "admin")
     expect(user.admin?).to be true
   end

2. Run spec → Confirm failure: bundle exec rspec spec/models/user_spec.rb (Fails with NoMethodError: undefined method 'admin?')
3. Implement (minimal):

   # app/models/user.rb
   def admin?
     role == "admin"
   end

4. Run spec → Confirm pass: bundle exec rspec spec/models/user_spec.rb (Passes)

Example: Bug fix with verification

Task: "Fix 500 error on Search when query is nil"

☐ Reproduce: Add a test case to spec/requests/search_spec.rb passing q: nil. ☐ Verify Failure: Run the spec and see it crash in SearchService. ☐ Fix: Add a null-object or guard in SearchService: query.to_s.strip. ☐ Verify Pass: Re-run the spec. ☐ Regression check: Run all search-related specs.

Rules

• Explicit Failure: Handle failure paths explicitly instead of leaving them implicit.
• Honest Verification: If a critical check could not be run, state it clearly.