| name | async-optimizer |
| description | Optimize async/concurrent code for throughput and correctness. Invoke for: "async issue", "concurrent", "race condition", "await optimization", "run in parallel", "asyncio", "Promise.all", "thread safety", "deadlock", "async bottleneck", "too many awaits", "sequential when should be parallel".
|
| argument-hint | async code file or function to optimize |
| allowed-tools | Read, Edit, Grep, Glob |
Skill: Async Optimizer — Concurrency & Parallelism
Category: Development
Role
Identify async antipatterns and optimize for maximum concurrency while maintaining correctness.
When to invoke
- Sequential awaits that could run in parallel
- Race conditions in async code
- Deadlocks or starvation
- "this async code is slow"
Instructions
- Read all async code in scope
- Find sequential awaits that are independent (can run with Promise.all / asyncio.gather)
- Identify race conditions: shared mutable state accessed concurrently?
- Check locks: proper async locks used? No blocking calls in async context?
- Find: CPU-bound work blocking event loop (needs thread pool)
- Optimize: gather independent tasks, use semaphores to limit concurrency
Output format
## Async Optimization — <file> — <date>
### Pattern Found: Sequential awaits (can be parallelized)
Before: [code]
After: [code using asyncio.gather()]
Speedup: ~3x for 3 independent operations
### Race Conditions
### Deadlock Risks
Example
/async-optimizer src/llm/claude_client.py — parallelize the batch completion calls
