Beyond your usual circle.
Asking friends by default can encourage rubber-stamping. Choose people for their knowledge of the changed code, with a clear reason to inspect it.
Find the people behind the code.
The owner. The author. The person who fixed it last time.
Find the right expertise, see the workload, and decide who to ask.
billing/refunds.pyRefund calculations
payments/retry.pyRetry behavior
tests/test_refunds.pyRegression coverage
OWNERS.md + contribution history
Owns the refund module and fixed its last rounding regression.
Ownership + prior bug fixRefund totals, currency rounding, and retry behavior.
Relevant context. A lighter queue.
Recommended, not requested. The final choice stays with you.
A familiar name can get you a quick approval.
The right
expertise brings meaningful scrutiny.
Asking friends by default can encourage rubber-stamping. Choose people for their knowledge of the changed code, with a clear reason to inspect it.
Sending every PR to the tech lead or senior architect creates a bottleneck. Find qualified teammates with more capacity instead of leaving their expertise unused.
Reconciling git blame, the PR diff, and past discussions is hard. CloudAEye connects those patterns with ownership and review queues to surface relevant candidates.
Poor reviews can let regressions reach production, damage customer trust, and contribute to churn. Match each change to people who can challenge it.
Reviewer selection · For developers opening a pull request
Recommend people by ownership, code familiarity, and contributions. See why each person fits and what they should inspect, not just a name on an ownership list.
Domain expertise · For sensitive and production-critical code
Match security-sensitive logic, critical services, shared components, and complex flows to people with relevant ownership and experience.
Prior contributions · For changes with hidden context
Find the developers who wrote, reviewed, or fixed the affected code. Their experience can help uncover subtle bugs, regressions, and design inconsistencies.
Review workload · For teams with overloaded leads
Prefer qualified reviewers with lighter queues. Distribute work without sending every PR to the same tech lead or senior architect.
Multi-component coverage · For changes across team boundaries
Map affected files, components, and responsibilities to complementary reviewers. Make clear who should inspect each area and the behavior between them.
Claude Code plugin · For repositories missing ownership context
Use git history to generate an initial root
OWNERS.md in your local working tree, then use it
for recommendations. The hosted PR integration does not create
this file in your checkout.
OWNERS.md
Review preparation · For developers piecing together the evidence
Connect git blame, the PR diff, ownership rules, and earlier PR discussions without a manual search. Bring expertise and workload together so the review can get moving sooner.
See the connected contextExplicit confirmation · For every review request
Read the rationale and inspection focus, then consider the team's circumstances. Reviews are requested only after you explicitly name the people. Appearing in a recommendation is not permission to send a request.
See the confirmation workflowA recommendation is the start of a decision, not permission to act on your behalf.
Start with the change and its affected code.
Compare the fit, review focus, and current queue.
Reviews are requested only after your explicit choice.
@cloudaeye /assign_reviewer
Maya brings refund ownership. Jon brings context on the retry path.
Illustrative preview only. No review request is sent.
Start in Claude Code or on your GitHub pull request.
/cloudaeye:assign-reviewer #412
@cloudaeye /assign_reviewer
The Claude Code workflow can create an initial
OWNERS.md at the repository root using git history.
That file stays in your local working tree, with no automatic
commit or push. The hosted PR integration does not create files in
your local checkout.
The GitHub App workflow requires repository write access and the
enabled Assign Reviewer feature. Token-only, GitLab, and Bitbucket
connections are not supported for this workflow. After choosing
reviewers, confirm with
@cloudaeye /assign_reviewer --apply @alice @bob,
using chosen names from the latest recommendation on this PR.
No. It recommends people and the areas they should inspect. It can reuse existing risk, slop, and review evidence rather than running those analyses again. The selected people still perform the review.
Check the PR before submitting another confirmed request. An interrupted job does not automatically resend the write, and queue completion is not proof that reviews were requested.