Multi-subagent code review across security, quality, performance, and architecture
Comparison
Code-quality (code-review) and deep-review are both
SDK-native review workflows reached with attune workflow run <slug>, both async, both predictive (LLM judgment) — but they
trade a different mix of passes and one has a narrowing knob.
code-quality (code-review) |
deep-review |
|
|---|---|---|
| Passes | Four: security, quality, performance, architecture | Three: security, quality, test gaps |
| Subagents | security- / quality- / perf- / architect-reviewer |
security- / quality- / test-gap-reviewer |
| Narrowing | None — scope via path only |
focus selects a subset of passes |
| Turn budget | 10 / 20 / 40 (quick / standard / deep) | 15 / 30 / 50 |
| Report sections | Summary / Security / Quality / Performance / Architecture / Suggestions | Summary / Security / Quality / Test Gaps / Suggestions |
| Slug | attune workflow run code-review |
attune workflow run deep-review |
Reach for code-quality for the everyday maintainability read —
the performance and architecture passes catch coupling and
inefficiency that a test-focused review skips. Reach for
deep-review when test coverage is the concern, or when you want
to narrow to a single domain with focus. The /code-quality
skill's "deep" option runs deep-review for exactly this reason.