Review Team
Review AI-written code before you commit it.
Code that runs on the happy path is the easiest kind to ship by mistake. The Review Team looks for the inputs that would break it.
Sample Thread
What the Review Team caught.
Write a Python function that removes duplicate users by email.
Here's a simple approach that keeps the first record for each email:
def dedupe(users):
seen = set()
result = []
for user in users:
if user["email"] not in seen:
seen.add(user["email"])
result.append(user)
return resultNormalize emails with strip() and lower() before comparing, skip records without an email, and add tests for those cases.
The code works on clean data, which is exactly why this bug would ship.
Real user tables have gaps. One bad row shouldn't take down the whole job.
Tests turn the review into something you can check, not just trust.
Sample thread written to show how the team works. Real comments depend on your reply.
Try It
Three steps on your own reply.
- Ask for the code
Have Claude or ChatGPT write the function as usual.
- Open the Review Team
Click ✦ Ask Alive under the reply and pick Review.
- Adopt and ask for a fix
Adopt the comments into the input box and ask the AI to revise the code.
More Use Cases
Review Team
Co-create Team
Fan TeamReview your next snippet.
Free Chrome extension for claude.ai and chatgpt.com. No account or API key needed.
Emails are compared as-is, so “Ana@site.com” and “ana@site.com ” survive as two users.
A record with no email field raises KeyError and stops the whole run.
Add tests for mixed case, trailing spaces, and a missing email before you merge.