mobiai-review-code

Review Code

Perform a mobile-specific code review, checking for common issues that static analysis tools miss.

When to Use

• User asks for a code review of mobile code changes
• After applying a fix, to self-review before creating a PR
• User shares a diff or PR and asks for feedback

Pre-flight: check the Graph for code structure

The Graph indexes Kotlin/Swift symbols across the repo, so callers of any symbol modified in the diff can be located instantly. Used during Step 3 (Review Checklist) and Step 4 (Verify Test Coverage) to detect call sites and tests that the diff may have missed.

If <repo>/.mobiai/graph/index.json exists:

# Freshness — if "hace Xd" with X ≥ 1, suggest refresh (don't auto-run):
mobiai graph status

# For each symbol modified/renamed in the diff, find callers outside the diff:
mobiai graph callers <SymbolFromDiff>

# If a symbol was renamed or removed, search for stale references:
mobiai graph search <OldSymbolName>

The callers list surfaces call sites that the diff may not have updated — a common source of silent breakage when the PR is partial.

If the index is ≥1 day old:

"El índice del Graph es de hace Xd. Si la rama tiene cambios recientes, considerá mobiai graph init antes de revisar."

Don't run init autonomously.

If .mobiai/graph/index.json does not exist and the project has .kt or .swift files:

"No veo el índice del Graph. Generalo con mobiai graph init para que pueda detectar callers fuera del diff."

Don't run init autonomously.

Skip silently when:

• The project has no Kotlin/Swift files (Flutter or RN-only).
• The diff is purely a config / resource / build-file change.

Workflow

Step 1: Identify the Correct Base Branch

<STOP> You MUST complete this step before reading ANY code or generating ANY diff. Do NOT skip this. Do NOT assume main or master. Get the real target branch first. </STOP>

Determine the target branch this change will be merged into:

1. If reviewing a PR: gh pr view --json baseRefName --jq '.baseRefName' — this is the authoritative answer
2. If the user mentioned a branch: use that branch
3. If no PR and no branch mentioned: ask the user "What branch is this being merged into?"
4. Only as absolute last resort: default to the repo's default branch via gh repo view --json defaultBranchRef --jq '.defaultBranchRef.name'

Then generate the diff against that branch:

BASE_BRANCH="<the branch from above>"
git diff "$BASE_BRANCH"...HEAD

Why this matters: If you diff against the wrong branch, you'll report hundreds of false changes that belong to the base branch, not to this PR. The user will see a noisy, useless review.

Step 2: Understand the Change

1. Read the diff against the correct base branch (from Step 1)
2. Understand the intent — what problem does this solve?
3. Check the related issue/ticket for context

Step 3: Review Checklist

Universal Checks

• [ ] Minimal change: Does the diff only contain what's necessary?
• [ ] No unrelated changes: No formatting, renaming, or refactoring mixed in?
• [ ] Error handling: Are errors handled appropriately (not swallowed silently)?
• [ ] Thread safety: Is shared mutable state properly synchronized?
• [ ] Null safety: Are nullable values handled correctly?
• [ ] Edge cases: Empty collections, null inputs, boundary values?

Android-Specific

• [ ] Lifecycle awareness: Are observers/listeners registered and unregistered properly?
• [ ] Memory leaks: No Activity/Fragment/Context references held in long-lived objects?
• [ ] Main thread: No blocking operations (network, disk I/O) on the main thread?
• [ ] Configuration changes: Does the code survive rotation/dark mode changes?
• [ ] ProGuard/R8: Are keep rules needed for reflection or serialization?
• [ ] Backward compatibility: Uses @RequiresApi or checks Build.VERSION.SDK_INT?

iOS-Specific

• [ ] Retain cycles: No strong reference cycles in closures? Using [weak self] where needed?
• [ ] Main thread UI: All UI updates on the main thread?
• [ ] Optional handling: No force unwraps (!) on values that could be nil?
• [ ] Memory: No large allocations in tight loops?
• [ ] App lifecycle: Handles sceneDidEnterBackground / applicationWillTerminate?

Flutter-Specific

• [ ] Widget rebuilds: No expensive operations in build() methods?
• [ ] State management: Proper use of the chosen state management solution?
• [ ] Platform channels: Error handling for platform-specific calls?
• [ ] Dispose: Controllers, streams, subscriptions disposed properly?

React Native-Specific

• [ ] Re-renders: Proper use of useMemo, useCallback, React.memo?
• [ ] Native bridges: Error handling for native module calls?
• [ ] Navigation: Proper cleanup in useEffect return?
• [ ] Platform-specific code: Platform.OS checks where needed?

Step 4: Verify Test Coverage

Check if the changed code has tests. This is a separate step — do not skip it.

1. Look at the diff: identify new logic, bug fixes, or behavior changes
2. Search for existing tests related to the modified files:
   • Android: look for test classes in src/test/ or src/androidTest/ that test the same classes
   • iOS: look for test files in the test target that cover the same types
   • Flutter: look for _test.dart files matching the changed files
   • React Native: look for .test.js / .test.tsx files
3. If there are NO tests for the changed code, report it as a blocking issue in your review:
   • "No unit tests found for [changed class/function]. This [bug fix / new logic / behavior change] should have test coverage."
4. If tests exist but don't cover the new changes, flag it as a suggestion

Always include a "Test coverage" section in your review output, even if tests are present. State what you found.

Step 5: Provide Feedback

Structure your review as:

1. Summary: One-line assessment (looks good / needs changes / has issues)
2. Issues found: Specific problems with file paths and line numbers
3. Suggestions: Optional improvements (clearly marked as non-blocking)

Handling User Feedback on Your Review

When the user responds to your review:

• Simple edits (README, changelog, docs, test adjustments) → make the change directly
• Code changes in the app → re-analyze with their feedback as context, apply the fix
• Questions → answer with specific code references
• Approval → done