The Review & Edit screen

When generation finishes, the wizard advances to Review & Edit — where you read what the AI produced, fix anything that needs polish, approve the instructions, and run the autograder check. This is your last gate before deploying to GitHub Classroom.

The four tabs

Each tab opens an in-browser editor for one of the four artifacts. Every change auto-saves as you type — no Save button to remember.

• Starter — what students will see. Has the TODO markers.
• Solution — the working answer. Lives on a separate solution branch in GitHub; never visible to students.
• Tests — the autograder rules: input/output pairs with point values.
• Instructions — the Markdown document students read (Instructions.md). Has an extra UI element — the approval gate banner at the top.

The progress bar

Above the tabs is the three-stage progress bar: Seed → Review & Edit → Deploy. The current stage is highlighted.

Click any completed stage to jump back. Common reasons:

• Click Seed to see (and edit) the wizard form inputs you originally submitted — useful if you want to change difficulty or learning objectives and regenerate.
• Click a deploy step to verify post-deploy state without losing your edits.

What to look at on each tab

Starter tab

This is what students see when they accept the assignment. Read it like a student would.

What to check:

• Are TODO markers clear? Each should describe what to implement in plain English, not just # TODO.
• Are function signatures preserved if students need to match a contract (e.g., for grading)?
• Is the entry point intact? Whatever runs the program should still be there.
• Any scaffolding code students shouldn't change? Add a comment like # Do not modify below this line if so.

What you can do:

• Edit any file directly in the Monaco editor (syntax highlighting, autocomplete, the works).
• Add new files with the + button on the file tab bar.
• Delete files with the × that appears on hover (but not the only file or the entry-point file).
• Click Regenerate in the top right to re-run the AI on just this artifact (uses one credit).

Solution tab

The AI's working answer. Same editor, same controls.

What to check:

• Does the solution actually solve the problem? Trace through one or two test cases mentally.
• Are there bugs? The autograder uses this as the source of truth — every test that passes against it codifies that buggy behaviour.
• Is the code at the right complexity for the difficulty level you picked? Beginner assignments shouldn't have list comprehensions and lambdas if students will see the solution post-grading.
• Does the solution match what the starter's TODOs are asking for? It's possible (rare) for the AI to drift between the two.

After significant solution edits, re-run validation so the autograder picks up the new behaviour.

Tests tab

The autograder rules. See Editing tests for the deep dive.

What to check:

• Coverage of edge cases (empty input, single value, large input, malformed input)
• Reasonable point allocations (typically 5–25 per test, totalling 100)
• Tests that actually pass against your solution — the validation step catches this, but eyeballing helps catch obvious typos

What you can do:

• Each test row has fields for Name, Comparison type (Exact match / Output contains / Regex), Input (stdin), Expected Output (stdout), and Points.
• Reorder by drag handle. Add by Add Test button. Delete with the trash icon.
• The total points across all tests is what shows up in GitHub Classroom's grading UI.
• Regenerate at the top re-runs the AI on the test set (replaces all current tests).

Instructions tab — has the approval gate

The student-facing assignment document. The yellow banner at the top is the approval gate — until you click Approve instructions, the autograder check is locked. See Approving instructions — the approval gate for what to actually look for during review.

Edit / Preview toggle: flip to Preview to see the rendered Markdown the way students will see it.

Regenerate in the top right re-runs only the Instructions.md generation at the current Difficulty — handy if you change the level after initial generation.

The Regenerate button — when to use it

Each tab has its own Regenerate button. Each one re-runs only that single artifact:

• Regenerate Starter — generates a new starter (with new TODOs) from your solution + difficulty
• Regenerate Solution — generates a new solution from your starter
• Regenerate Tests — generates a fresh test set against your solution (replaces all current tests)
• Regenerate Instructions — re-runs Instructions.md at the current difficulty

Each regeneration costs one credit. Use sparingly — if you're not happy with what came back, often editing in place is faster than another regeneration roll. The AI doesn't always produce a meaningfully different result on a second roll.

What to do when you're done reviewing

1. Read all four tabs end to end.
2. Edit anything that needs polish (typos, tweaks, extra tests).
3. Open the Instructions tab → click Approve instructions in the yellow banner.
4. The yellow banner turns green and a new Run autograder check button appears.
5. Click Run autograder check to start validation.

See The autograder check for what happens next.

Status flow on this screen

While you're on Review & Edit, the assignment status moves through:

• generated — generation finished, you're reviewing
• validating — you clicked Run autograder check, the sandbox is running
• validated — autograder passed 100%, ready for Deploy
• back to generated (with a reviewIntervention flag) — autograder failed, you're back to fixing it; see When validation fails

Until status is validated, the Continue to Deploy button stays locked.

Where to go next

• Ready to validate? → The autograder check explains what runs during validation and what passes / fails actually mean.
• Want to add more tests before validating? → Editing tests.
• Approval gate banner is unclear? → Approving instructions — the approval gate.
• Validation already failed? → When validation fails.