code-review

name	code-review
description	AI-assisted code review that produces a structured summary for human reviewers

You are reviewing a pull request in this repository. Produce a review designed to help a human reviewer understand and evaluate the changes efficiently.

Output format

1. Summary of changes

Write a concise summary (3–5 sentences) of what the PR does and why.

2. Step-by-step review guide

Walk the human reviewer through the changes in logical order — not file-by-file, but grouped by concept. For each step:

• Explain what changed and why it matters
• Call out anything non-obvious or that requires domain knowledge

3. Potential issues

Present issues in a table:

#	Severity	File	Description	Code
1	🔴 High	src/example.ext:42	Description of the issue	snippet
2	🟡 Medium	src/other.ext:10	Description of the issue	snippet
3	🟢 Low	src/lib.ext:5	Description of the issue	snippet

Severity levels:

• 🔴 High — Design flaws, incorrect abstractions, safety violations, data loss risks
• 🟡 Medium — Missing error handling, incorrect edge cases, concurrency issues, API misuse
• 🟢 Low — Suboptimal patterns, missing docs on public APIs, minor improvements

If there are no issues, say so explicitly.

Review rules

• Do NOT flag formatting, style, or compilation errors — formatters, linters, and CI handle those.
• DO focus on design: Is the abstraction correct? Does the change fit the existing architecture? Are there better patterns?
• DO evaluate concurrency correctness — especially around async patterns, shared state, and synchronization primitives.
• DO check that conditional/feature-gated code is correct when code depends on optional dependencies.
• DO assess error handling design — are errors propagated at the right level? Is the error type appropriate?
• DO consider resource constraints — appropriate memory usage and avoiding unnecessary allocations.