Project 01: Should both starter and solution versions result in a fully functional Electron app?

[dauvannam1804]

Description

Hi,

I'm currently working through Project 01 ("Prompt-Only vs. Rules-First: How Much Difference Does It Make") and would like to clarify the expected success criteria.

What I did

For the starter/ version:

Changed into the starter/ directory.
Provided the contents of task-prompt.md to my coding agent (Antigravity).
Waited for the agent to finish implementation.
Ran the application with:
npm run dev

The Electron window launches successfully, and the UI is visible. However, the application is mostly non-functional (document list, Q&A interactions, etc. do not appear to work as expected).

Evidence (starter)

For the solution/ version:

Changed into the solution/ directory.
Provided the task prompt to the same coding agent.
Waited for implementation to complete.
Ran:
npm run dev

Again, the Electron window launches and displays a UI, but the expected functionality does not seem to be working.

Evidence (solution)

Question

For Project 01, what is the intended completion criteria?

Should the outcome be:

Option A

The exercise is primarily about comparing agent behavior.
It is acceptable if the generated applications are incomplete or partially functional.
The focus is on observing differences between prompt-only and harness-assisted runs.

Option B

Both starter and solution runs are expected to produce a working Electron application.
The document list, Q&A panel, and local data directory functionality should all work correctly after the agent completes the task.
If they do not work, that indicates the run was not completed successfully.

I want to make sure I'm evaluating the project correctly and not misunderstanding the intended goal of the exercise.

Thanks!

[georgejdli]

I was initially confused as well but I believe it's Option A since the subsequent projects built on the previous project code.

I noticed that the src files in both the starter and the solution looked the same. My approach to Project 01 was to delete the src folder in the starter folder and prompt my agent to complete the task without looking at other directories in this repository to see how the weak harness would perform from scratch. Then I did the same in the solution directory, deleted the CLAUDE.MD, and reset the claude-progress.md with the template to see how my agent performed with a strong harness.