<skill_overview> <purpose>Keep Rust code fast while staying idiomatic</purpose> <triggers> <trigger>Optimizing hot loops</trigger> <trigger>Reducing allocations</trigger> <trigger>Choosing data structures</trigger> <trigger>Measuring performance impact</trigger> </triggers> <sources> <source url="https://doc.rust-lang.org/book/ch13-04-performance.html">The Rust Book - Performance</source> </sources> </skill_overview> <zero_cost_abstractions> <rules> <rule>Use iterators and closures freely; they compile to efficient code</rule> <rule>Prefer expressive, safe code and measure before micro-optimizing</rule> </rules> </zero_cost_abstractions> <allocations> <rules> <rule>Prefer borrowing (&str, &[T]) over allocating (String, Vec)</rule> <rule>Pre-allocate with Vec::with_capacity when size is known</rule> <rule>Avoid unnecessary clones; pass references where possible</rule> </rules> <example> <code> let mut buf = Vec::with_capacity(1024); </code> </example> </allocations> <hot_paths> <rules> <rule>Avoid repeated formatting or allocation in tight loops</rule> <rule>Reuse buffers and temporary values when possible</rule> </rules> </hot_paths> <measurement> <rules> <rule>Benchmark with realistic data sizes</rule> <rule>Compare before and after changes</rule> </rules> </measurement> <anti_patterns> <avoid name="premature_optimization">Optimizing before profiling</avoid> <avoid name="clone_in_loops">Cloning on each iteration</avoid> <avoid name="unbounded_growth">Growing collections without bounds</avoid> </anti_patterns>