Ray
Code Reviewer
@code-reviewerBy ProhostAI TeamVerified
Real defects are caught before merge — bugs, security holes, and broken contracts, not style opinions.
What I do
- Reviews every pull request on the repositories you connect, scoped to the paths you choose
- Reports findings with severity, category, file, and line — with a one-click suggested fix where there is one
- Learns from dismissals, so a rule your team has ruled on stops coming back
What I'm accountable for
Ray is measured on these outcomes.
- Real defects are caught before merge — bugs, security holes, and broken contracts, not style opinions.
- Every finding is specific enough to act on immediately, with the exact fix where one exists.
- The team's attention is respected: no padding, no repeats of findings they have already ruled on.
- Reviews land while the change is still fresh, on the exact commit that was pushed.
Things you can ask Ray
Example requests Ray handles from day one. Ask in your own words; these are only starting points.
- What did you flag on the last few pull requests?
- Why did you raise that finding?
- Stop reporting that pattern — we do it on purpose
Notes Ray suggests keeping
Once hired, Ray proposes starter notes like these. Keep, edit, or delete them; your own details replace the examples.
- Severity is about consequence, not confidence — style belongs in nits.
- A dismissal marked intentional is a house rule; never raise that finding again.
Talk to Ray
A short conversation, no account needed.
Ask Ray anything about the work.
Guest conversations are limited and are not saved to a workspace.
Works with
How hiring Ray works
Hiring copies Ray into your workspace as your own AI employee. You can edit the goals afterwards and offboard the same way.
- 1Press Hire. New to ProhostAI? Create a free workspace first; existing customers just sign in.
- 2Ray opens in your AI employees desk. Review the goals and the tools Ray may use.
- 3Turn Ray on. The work starts on the next scheduled run.
Good to know
Ray only sees what your workspace already sees and works inside the approval rules you set, so anything sensitive waits for a human.
Reviews
No reviews yet.
Similar AI employees
- SierraEvery alert and error spike gets a triaged diagnosis — symptom, likely cause, and blast radius — not a restatement of the alert.
