| description | Use when building Rust systems where memory safety, ownership patterns, zero-cost abstractions, and performance optimization are critical for systems programming, embedded development, async applications, or high-performance services. Specifically:\n\n<example>\nContext: Designing a high-throughput async network service that must handle thousands of concurrent connections with minimal memory overhead and strict latency requirements\nuser: "Create a tokio-based async service that processes 50k concurrent TCP connections. Need proper ownership patterns to avoid allocations, async/await with custom Future implementations where needed, and memory-safe FFI bindings to a C library. Must be zero-unsafe-code in public API."\nassistant: "I'll architect the service with smart pointer patterns (Arc for shared state, Box for heap allocation), async task spawning with cancellation via select!, proper lifetime management for the FFI boundary, and unsafe blocks only in isolated wrapper crates with exhaustive MIRI verification. This ensures sub-microsecond latency with predictable memory usage."\n<commentary>\nUse rust-engineer when building async/concurrent systems with tight performance budgets, strict memory constraints, or complex ownership patterns that need careful lifetime management to achieve zero-allocation paths in hot code.\n</commentary>\n</example>\n\n<example>\nContext: Implementing a systems programming project like a file parser, codec library, or embedded driver with complex unsafe code requirements and memory safety concerns\nuser: "Building a binary file parser that must handle untrusted input safely and efficiently. Need custom allocators for arena allocation, unsafe code for SIMD optimizations, careful bounds checking, and comprehensive testing with MIRI to catch undefined behavior. Should compile to both x86_64 and ARM targets."\nassistant: "I'll design the parser with safe abstractions over unsafe code blocks, use custom Allocator trait for arena patterns, implement SIMD intrinsics safely within isolated unsafe modules, validate all invariants, add fuzzing with cargo-fuzz, verify with MIRI, and ensure clippy::pedantic passes. Document all safety invariants thoroughly."\n<commentary>\nInvoke rust-engineer for systems-level code that requires unsafe blocks, custom memory management, SIMD intrinsics, embedded constraints (no_std), or cross-platform compilation where memory safety verification is non-negotiable.\n</commentary>\n</example>\n\n<example>\nContext: Performance optimization for an existing Rust codebase hitting memory/CPU limits with profiling data indicating allocation hotspots and GC pressure\nuser: "Our parser is allocating 50MB per request. Profile shows most allocations in String building and Vec resizing. Need to apply Cow patterns, use custom types with SmallVec for stack allocation, benchmark against current implementation, and document the optimization tradeoffs."\nassistant: "I'll apply profiling with flamegraph, identify hot paths, replace allocating patterns with Cow<str> and SmallVec<[T; N]>, implement custom iterators to reduce intermediate allocations, add criterion benchmarks showing improvements, and verify with perf that cache behavior improves. Zero-allocation paths for critical code."\n<commentary>\nUse rust-engineer for performance-critical optimization work, benchmarking against baselines, zero-allocation optimizations, memory-efficient data structures, or when Rust's type system needs to encode performance guarantees at compile-time.\n</commentary>\n</example> |
