Rust Best Practices

Comprehensive guide for writing high-quality, idiomatic, and highly optimized Rust code. Contains 193 rules across 14 categories, prioritized by impact to guide LLMs in code generation and refactoring. Quick reference below; the reasoning behind each rule family lives in the per-source digests under references/ (adapted from BeautifulRust, MIT).

When to Apply

Reference these guidelines when:

• Writing new Rust functions, structs, or modules
• Implementing error handling or async code
• Designing public APIs for libraries
• Reviewing code for ownership/borrowing issues
• Optimizing memory usage or reducing allocations
• Tuning performance for hot paths
• Refactoring existing Rust code

Rule Categories by Priority

Quick Reference

1. Ownership & Borrowing (CRITICAL)

• own-borrow-over-clone - Prefer &T borrowing over .clone()
• own-slice-over-vec - Accept &[T] not &Vec<T>, &str not &String
• own-cow-conditional - Use Cow<'a, T> only for real conditional ownership
• own-arc-shared - Use Arc<T> for thread-safe shared ownership
• own-rc-single-thread - Use Rc<T> for single-threaded sharing
• own-refcell-interior - Use RefCell<T> for interior mutability (single-thread)
• own-mutex-interior - Use Mutex<T> for interior mutability (multi-thread)
• own-rwlock-readers - Use RwLock<T> when reads dominate writes
• own-copy-small - Derive Copy for small, trivial types
• own-clone-explicit - Make Clone explicit, avoid implicit copies
• own-move-large - Move large data instead of cloning
• own-lifetime-elision - Rely on lifetime elision when possible
• own-borrow-counter-ladder - Fix borrow errors by moving data: mem::take, struct splitting, typed intermediates
• own-drop-contract - Destructors never fail, panic, or block; bind guards to names

2. Error Handling (CRITICAL)

• err-thiserror-lib - Use thiserror for library error types
• err-anyhow-app - Use anyhow for application error handling
• err-result-over-panic - Return Result, don't panic on expected errors
• err-context-chain - Add context with .context() or .with_context()
• err-no-unwrap-prod - Never use .unwrap() in production code
• err-expect-bugs-only - Use .expect() only for programming errors
• err-question-mark - Use ? operator for clean propagation
• err-from-impl - Use #[from] for automatic error conversion
• err-source-chain - Use #[source] to chain underlying errors
• err-lowercase-msg - Error messages: lowercase, no trailing punctuation
• err-doc-errors - Document errors with # Errors section
• err-custom-type - Create custom error types, not Box<dyn Error>
• err-panic-policy - Panics are not exceptions; poisoned locks are fatal; catch_unwind is last resort
• err-app-lib-boundary - One app error type; libraries never expose anyhow; return consumed args in errors

3. Memory Optimization (CRITICAL)

• mem-with-capacity - Use with_capacity() when size is known
• mem-smallvec - Use SmallVec for usually-small collections
• mem-arrayvec - Use ArrayVec for bounded-size collections
• mem-box-large-variant - Box large enum variants to reduce type size
• mem-boxed-slice - Use Box<[T]> instead of Vec<T> when fixed
• mem-thinvec - Use ThinVec for often-empty vectors
• mem-clone-from - Use clone_from() to reuse allocations
• mem-reuse-collections - Reuse collections with clear() in loops
• mem-avoid-format - Avoid format!() when string literals work
• mem-write-over-format - Use write!() instead of format!()
• mem-arena-allocator - Use arena allocators for batch allocations
• mem-zero-copy - Use zero-copy patterns with slices and Bytes
• mem-compact-string - Use CompactString for small string optimization
• mem-smaller-integers - Use smallest integer type that fits
• mem-assert-type-size - Assert hot type sizes to prevent regressions

4. API Design (HIGH)

• api-builder-pattern - Use Builder pattern for complex construction
• api-builder-must-use - Add #[must_use] to builder types
• api-newtype-safety - Use newtypes for type-safe distinctions
• api-typestate - Use typestate for compile-time state machines
• api-sealed-trait - Seal traits to prevent external implementations
• api-extension-trait - Use extension traits to add methods to foreign types
• api-parse-dont-validate - Parse into validated types at boundaries
• api-impl-into - Accept impl Into<T> for flexible string inputs
• api-impl-asref - Accept impl AsRef<T> for borrowed inputs
• api-must-use - Add #[must_use] to Result returning functions
• api-non-exhaustive - Use #[non_exhaustive] for future-proof enums/structs
• api-from-not-into - Implement From, not Into (auto-derived)
• api-default-impl - Implement Default for sensible defaults
• api-common-traits - Implement Debug, Clone, PartialEq eagerly
• api-serde-optional - Gate Serialize/Deserialize behind feature flag
• api-no-bool-params - No bool parameters in public signatures - enum or two methods
• api-ffi-boundary - FFI: repr(C), same-side alloc/free, RAII pointer wrappers, foo-sys/bindgen

5. Async/Await (HIGH)

• async-tokio-runtime - Use Tokio for production async runtime
• async-no-lock-await - Never hold Mutex/RwLock across .await
• async-spawn-blocking - Use spawn_blocking for CPU-intensive work
• async-tokio-fs - Use tokio::fs not std::fs in async code
• async-cancellation-token - Use CancellationToken for graceful shutdown
• async-join-parallel - Use tokio::join! for parallel operations
• async-try-join - Use tokio::try_join! for fallible parallel ops
• async-select-racing - Use tokio::select! for racing/timeouts
• async-bounded-channel - Use bounded channels for backpressure
• async-mpsc-queue - Use mpsc for work queues
• async-broadcast-pubsub - Use broadcast for pub/sub patterns
• async-watch-latest - Use watch for latest-value sharing
• async-oneshot-response - Use oneshot for request/response
• async-joinset-structured - Use JoinSet for dynamic task groups
• async-clone-before-await - Clone data before await, release locks
• async-send-sync-assert - Pin Send/Sync with compile-time assert_send checks

6. Compiler Optimization (HIGH)

• opt-inline-small - Use #[inline] for small hot functions
• opt-inline-always-rare - Use #[inline(always)] sparingly
• opt-inline-never-cold - Use #[inline(never)] for cold paths
• opt-cold-unlikely - Use #[cold] for error/unlikely paths
• opt-likely-hint - Use likely()/unlikely() for branch hints
• opt-lto-release - Enable LTO in release builds
• opt-codegen-units - Use codegen-units = 1 for max optimization
• opt-pgo-profile - Use PGO for production builds
• opt-target-cpu - Set target-cpu=native for local builds
• opt-bounds-check - Use iterators to avoid bounds checks
• opt-simd-portable - Use portable SIMD for data-parallel ops
• opt-cache-friendly - Design cache-friendly data layouts (SoA)

7. Naming Conventions (MEDIUM)

• name-types-camel - Use UpperCamelCase for types, traits, enums
• name-variants-camel - Use UpperCamelCase for enum variants
• name-funcs-snake - Use snake_case for functions, methods, modules
• name-consts-screaming - Use SCREAMING_SNAKE_CASE for constants/statics
• name-lifetime-short - Use short lowercase lifetimes: 'a, 'de, 'src
• name-type-param-single - Use single uppercase for type params: T, E, K, V
• name-as-free - as_ prefix: free reference conversion
• name-to-expensive - to_ prefix: expensive conversion
• name-into-ownership - into_ prefix: ownership transfer
• name-no-get-prefix - No get_ prefix for simple getters
• name-is-has-bool - Use is_, has_, can_ for boolean methods
• name-iter-convention - Use iter/iter_mut/into_iter for iterators
• name-iter-method - Name iterator methods consistently
• name-iter-type-match - Iterator type names match method
• name-acronym-word - Treat acronyms as words: Uuid not UUID
• name-crate-no-rs - Crate names: no -rs suffix
• name-weasel-types - No Service/Manager/Factory/Helper/Util; don't repeat the module name in the type

8. Type Safety (MEDIUM)

• type-newtype-ids - Wrap IDs in newtypes: UserId(u64)
• type-newtype-validated - Newtypes for validated data: Email, Url
• type-enum-states - Use enums for mutually exclusive states
• type-option-nullable - Use Option<T> for nullable values
• type-result-fallible - Use Result<T, E> for fallible operations
• type-phantom-marker - Use PhantomData<T> for type-level markers
• type-never-diverge - Use ! type for functions that never return
• type-generic-bounds - Add trait bounds only where needed
• type-no-stringly - Avoid stringly-typed APIs, use enums/newtypes
• type-repr-transparent - Use #[repr(transparent)] for FFI newtypes
• type-strongest-std - Strongest std types: PathBuf, Duration, NonZeroU32 internal-only, bitflags vs enum

9. Testing (MEDIUM)

• test-cfg-test-module - Use #[cfg(test)] mod tests { }
• test-use-super - Use use super::*; in test modules
• test-integration-dir - Put integration tests in tests/ directory
• test-descriptive-names - Use descriptive test names
• test-arrange-act-assert - Structure tests as arrange/act/assert
• test-proptest-properties - Use proptest for property-based testing
• test-mockall-mocking - Use mockall for trait mocking
• test-mock-traits - Use traits for dependencies to enable mocking
• test-fixture-raii - Use RAII pattern (Drop) for test cleanup
• test-tokio-async - Use #[tokio::test] for async tests
• test-should-panic - Use #[should_panic] for panic tests
• test-criterion-bench - Use criterion for benchmarking
• test-doctest-examples - Keep doc examples as executable tests
• test-refactor-pin-behavior - Pin current behavior with tests before refactoring; list signature changes
• test-miri-fuzz - Miri over unsafe; cargo fuzz untrusted-input parsers

10. Documentation (MEDIUM)

• doc-all-public - Document all public items with ///
• doc-module-inner - Use //! for module-level documentation
• doc-examples-section - Include # Examples with runnable code
• doc-errors-section - Include # Errors for fallible functions
• doc-panics-section - Include # Panics for panicking functions
• doc-safety-section - Include # Safety for unsafe functions
• doc-question-mark - Use ? in examples, not .unwrap()
• doc-hidden-setup - Use # prefix to hide example setup code
• doc-intra-links - Use intra-doc links: [Vec]
• doc-link-types - Link related types and functions in docs
• doc-cargo-metadata - Fill Cargo.toml metadata

11. Performance Patterns (MEDIUM)

• perf-iter-over-index - Prefer iterators over manual indexing
• perf-iter-lazy - Keep iterators lazy, collect() only when needed
• perf-collect-once - Don't collect() intermediate iterators
• perf-entry-api - Use entry() API for map insert-or-update
• perf-drain-reuse - Use drain() to reuse allocations
• perf-extend-batch - Use extend() for batch insertions
• perf-chain-avoid - Avoid chain() in hot loops
• perf-collect-into - Use collect_into() for reusing containers
• perf-black-box-bench - Use black_box() in benchmarks
• perf-release-profile - Optimize release profile settings
• perf-profile-first - Profile before optimizing

12. Project Structure (LOW)

• proj-lib-main-split - Keep main.rs minimal, logic in lib.rs
• proj-mod-by-feature - Organize modules by feature, not type
• proj-flat-small - Keep small projects flat
• proj-mod-rs-dir - Use mod.rs for multi-file modules
• proj-pub-crate-internal - Use pub(crate) for internal APIs
• proj-pub-super-parent - Use pub(super) for parent-only visibility
• proj-pub-use-reexport - Use pub use for clean public API
• proj-prelude-module - Create prelude module for common imports
• proj-bin-dir - Put multiple binaries in src/bin/
• proj-workspace-large - Use workspaces for large projects
• proj-workspace-deps - Use workspace dependency inheritance
• proj-semver-features - Semver is API: additive bare-named features, deprecate-then-remove, deliberate MSRV

13. Clippy & Linting (LOW)

• lint-deny-correctness - #![deny(clippy::correctness)]
• lint-warn-suspicious - #![warn(clippy::suspicious)]
• lint-warn-style - #![warn(clippy::style)]
• lint-warn-complexity - #![warn(clippy::complexity)]
• lint-warn-perf - #![warn(clippy::perf)]
• lint-pedantic-selective - Enable clippy::pedantic selectively
• lint-missing-docs - #![warn(missing_docs)]
• lint-unsafe-doc - #![warn(clippy::undocumented_unsafe_blocks)]
• lint-cargo-metadata - #![warn(clippy::cargo)] for published crates
• lint-rustfmt-check - Run cargo fmt --check in CI
• lint-workspace-lints - Configure lints at workspace level
• lint-expect-over-allow - #[expect(lint, reason)] over #[allow]; never #![deny(warnings)] in source

14. Anti-patterns (REFERENCE)

• anti-unwrap-abuse - Don't use .unwrap() in production code
• anti-expect-lazy - Don't use .expect() for recoverable errors
• anti-clone-excessive - Don't clone when borrowing works
• anti-lock-across-await - Don't hold locks across .await
• anti-string-for-str - Don't accept &String when &str works
• anti-vec-for-slice - Don't accept &Vec<T> when &[T] works
• anti-index-over-iter - Don't use indexing when iterators work
• anti-panic-expected - Don't panic on expected/recoverable errors
• anti-empty-catch - Don't use empty if let Err(_) = ... blocks
• anti-over-abstraction - Don't over-abstract with excessive generics
• anti-premature-optimize - Don't optimize before profiling
• anti-type-erasure - Don't use Box<dyn Trait> when impl Trait works
• anti-format-hot-path - Don't use format!() in hot paths
• anti-collect-intermediate - Don't collect() intermediate iterators
• anti-stringly-typed - Don't use strings for structured data
• anti-unsound-api - A safe API that can cause UB through safe calls is unsound

Deep Dive References

When you need the reasoning behind a rule family — or a complete checklist for a task like publishing a crate or auditing an API — read the per-source digests in references/ (adapted from BeautifulRust, MIT):

• references/rust-api-guidelines.md - all 54 official C-codes with the full naming, interoperability, and predictability checklist
• references/effective-rust.md - 35 items on choosing between alternatives: generics vs dyn, Result vs panic, iterators vs loops, borrow-checker tactics, dependency/CI policy
• references/rust-design-patterns.md - idioms, patterns (builder, RAII, newtype, visitor, mem::take), and anti-patterns with code
• references/microsoft-rust-guidelines.md - M-codes for teams: error struct design, panic/unsafe policy, workspace layout, agent-specific guidance
• references/google-rust-style-guide.md - formatting, imports, module layout, accessor naming, and the application-vs-library distinction

Note on conflicting guidance: one habit from this skill's older lint rules is superseded — never put #![deny(warnings)] in source (new toolchains add warnings that break your build). Configure lint levels once in Cargo.toml ([lints.rust] / [workspace.lints]), override narrowly with #[expect(lint, reason)], and enforce warnings-as-errors in CI with cargo clippy -- -D warnings. See lint-expect-over-allow.

Recommended Cargo.toml Settings

[profile.release]
opt-level = 3
lto = "fat"
codegen-units = 1
panic = "abort"
strip = true

[profile.bench]
inherits = "release"
debug = true
strip = false

[profile.dev]
opt-level = 0
debug = true

[profile.dev.package."*"]
opt-level = 3  # Optimize dependencies in dev

How to Use

This skill provides rule identifiers for quick reference. When generating or reviewing Rust code:

1. Check relevant category based on task type
2. Apply rules with matching prefix
3. Prioritize CRITICAL > HIGH > MEDIUM > LOW
4. Read rule files in rules/ for detailed examples

Rule Application by Task

PR Code Review

For automated PR bug review and fix loops, use the pr-code-review skill/agent. It ranks bugs critical/high/medium/low, eliminates false positives, applies KISS/DRY fixes (main logic only, skip tests), and re-reviews after every change. See AGENTS.md#pr-code-review-agent for full workflow.

/pr-code-review                    # review current uncommitted diff
/pr-code-review --branch main      # review branch diff against main
/pr-code-review --last-diff        # review the last committed diff

Sources

This skill synthesizes best practices from:

• Rust API Guidelines
• Rust Performance Book
• Google Rust Style Guide
• Microsoft Pragmatic Rust Guidelines
• Effective Rust (David Drysdale)
• Rust Design Patterns
• Production codebases: ripgrep, tokio, serde, polars, axum, deno
• Clippy lint documentation
• Community conventions (2024-2026)

v1.1.0 additions adapted from BeautifulRust by Morgan Linton (MIT License, © 2025), which synthesizes the Google Rust Style Guide, the Rust API Guidelines, Effective Rust, Rust Design Patterns, and Microsoft's Pragmatic Rust Guidelines; full per-source digests are in references/.