Reviewing a diff against a written checklist is the one job in this catalogue where a coding harness is doing what it was actually built for. Codex reads code fluently and is strong on the surrounding API work of pulling commits, pipeline status and existing notes. The output is a fixed four heading note, so what is being asked for is accuracy rather than opinion.
Run a staged conversation — no account needed. The agent handles it for real while a simulated world answers its tool calls; nothing touches real accounts, and nothing is actually sent.
Add idempotency keys to the refunds endpoint
Open@fatima.zahra
Adds an Idempotency-Key header to POST /refunds and stores the key against the refund for 24 hours, so a retried request returns the original refund instead of issuing a second one. Migration adds one table and a unique index. Twelve files, 340 lines added, 46 removed. Unit and integration tests included, pipeline green on the head commit. Closes issue 2841.
Using this template drops you into a guided setup. It asks exactly this, nothing else:
Connect GitLab
One sign-in. The agent acts through your account, scoped to what this template uses.
Connect Slack
One sign-in. The agent acts through your account, scoped to what this template uses.
Review checklist
What a merge request has to satisfy on this project in your own words, which of those items block a merge, and what you want asked of the author when an item cannot be judged from the diff alone.
Runs on Codex
Preselected for this page — connect your Codex account during setup, or switch to NoClick's built-in models with one click.
Watch it handle a test run
A staged conversation against a simulated world — then it’s live.
A checklist item about a missing migration note or an untested branch needs the change understood rather than summarised. This is native ground for the harness.
An item it cannot evidence from the diff is marked unclear with the reason given, never failed. That is what keeps authors trusting the checklist instead of arguing with it.
No. Each note opens with what moved since the last one and drops points the previous note already made, so a rapid series of pushes produces short follow ups rather than the same review repeated. Six pushes is still six notes, and teams who find that busy usually trim the checklist to blocking items, which shortens the follow ups further.
Free to start. Guided setup, a test run against staged conversations, and it's live.