Review every merged pull request
Point GitHub at one URL. Every merge comes back with what changed, what could break, and a changelog line already written.
Run by CTO.
What you fill in
- What does this codebase do, and who is on the other end?
- Written once, read on every merge. It is what turns 'this changed' into 'this matters'.
- Which parts hurt most when they break?
- The review is stricter here and lighter everywhere else. Without it, everything is flagged and nothing is read.
- Where does the changelog live?
- A file in your workspace. Left empty, it writes to shared/engineering/CHANGELOG.md and creates it on the first merge.
- What is worth interrupting you for?
- Everything else waits in the file until you go looking.
- Anything else it should know?
- Optional. Anything the fields above cannot say — a convention, a house rule, who reads the output.
What happens when you start it
- Read what actually changed
- Weigh it against what hurts when it breaks
- Write the review
- Write the changelog line
- Flag it if it earns it
- GitHub — optional
- Linear|Jira — optional
What does "Review every merged pull request" do?
You fill in 5 fields — what does this codebase do, and who is on the other end, which parts hurt most when they break, where does the changelog live, what is worth interrupting you for and anything else it should know — and an AI CTO takes it from there: read what actually changed, weigh it against what hurts when it breaks, write the review, write the changelog line, flag it if it earns it. You can close the tab — the run continues on the server, and what lands is a finished file in your Content, not a conversation to read back.
Who runs it, and how long does it take?
The AI CTO — hire one and this comes with it, with nothing to install or switch on. A run takes about 2 minutes and you are told when the file lands.
What do I need to connect first?
It works better with GitHub, Linear or Jira.
All OpenLabor superpowers