Menu
Rust code review patterns from corrode.dev (16 articles)
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
| name | idiomatic-rust |
| description | Rust code review patterns from corrode.dev (16 articles) |
| license | CC-BY |
| compatibility | opencode |
| metadata | {"source":"corrode.dev/blog","articles":"16"} |
This skill provides guidance on writing idiomatic, production-ready Rust code based on 16 comprehensive articles from corrode.dev by Matthias Endler. Each pattern includes a summary and link to the full article, which you can fetch using the webfetch tool when you need detailed context.
Avoid over-engineering and premature abstraction. The biggest mistake is making code generic "because we might need it someday." Use concrete types unless you're actively switching implementations. Generics increase compile time, complexity, and cognitive load. Simple code is easier to debug, maintain, and often performs better because the compiler can optimize it more effectively. Don't be clever—be clear.
📖 https://corrode.dev/blog/simple/
Make the compiler enforce your invariants instead of relying on runtime checks. Destructure structs in trait implementations to catch when fields are added. Use slice pattern matching instead of indexing. Avoid ..Default::default() because it hides which fields you're setting. Never use From implementations that can panic—use TryFrom instead. The goal is to make implicit assumptions explicit and compiler-checked.
📖 https://corrode.dev/blog/defensive-programming/
Memory safety doesn't mean your code can't have bugs. Check arithmetic with checked_mul, checked_add, etc. Never use as for narrowing conversions—use TryFrom. Don't index arrays directly—use .get() or slice patterns. Create newtype wrappers for domain concepts instead of using raw String or u32. Make illegal states unrepresentable with enums. Implement Debug carefully to avoid leaking secrets. Use constant-time comparison for passwords and sensitive data.
📖 https://corrode.dev/blog/pitfalls-of-safe-rust/
When prototyping, prioritize speed of iteration over perfection. Use liberal unwrap() and clone(). Don't worry about allocations. Accept performance crimes. The goal is to validate ideas quickly, then refactor with the knowledge you've gained. Rust is fast enough that your "slow" prototype is probably faster than other languages' production code anyway.
📖 https://corrode.dev/blog/prototyping/
The standard library has gotchas. Path::join with an absolute path replaces the base path entirely. split_at panics if the index is out of bounds—use split_at_checked. Some Default implementations create invalid states. Be aware that even safe Rust APIs can have surprising behavior.
📖 https://corrode.dev/blog/sharp-edges-in-rust-std/
Stop using unwrap() and expect() in production code. Use ? for propagation, ok_or() for conversion to Result, unwrap_or() for defaults, map() for transformation, and pattern matching for complex cases. Each approach communicates intent better than a blind unwrap. The only acceptable uses are in tests and when you've proven the invariant elsewhere in the code.
📖 https://corrode.dev/blog/rust-option-handling-best-practices/
Most of the time, you don't need explicit lifetimes. The compiler infers them correctly in 95% of cases. Only add lifetime annotations when the compiler can't figure out the relationship between references. Don't try to outsmart the borrow checker—listen to what it's telling you about your design. If you're fighting lifetimes, the problem is usually your API design, not the syntax.
📖 https://corrode.dev/blog/lifetimes/
Rust supports multiple paradigms—imperative, functional, and object-oriented. Use the right tool for the job. Iterators and pattern matching are functional. Structs with methods are OOP. Loops and mutation are imperative. Don't force one paradigm everywhere. The best Rust code mixes paradigms appropriately based on the problem being solved.
📖 https://corrode.dev/blog/paradigms/
Rust is expression-oriented. Use if, match, and blocks as expressions that return values. This leads to more declarative code with fewer intermediate variables and mutations. Instead of setting a variable in each branch, return the value from the expression. This style prevents uninitialized variables and makes data flow clearer.
📖 https://corrode.dev/blog/expressions/
Prefer let over let mut. Immutability prevents entire classes of bugs and makes code easier to reason about. When you need temporary mutability for initialization, shadow the variable to make it immutable afterward. Immutable data can be safely shared across threads. The compiler optimizes immutable code better because it knows the values can't change.
📖 https://corrode.dev/blog/immutability/
Replace loops with iterator chains when possible. Iterators are lazy, composable, and often compile to the same code as manual loops—but they're more declarative. Use filter, map, flat_map, take, skip, and other adapters to express intent clearly. The only time to use explicit loops is when the iterator version is genuinely less readable or when you need early returns.
📖 https://corrode.dev/blog/iterators/
Enums are more powerful in Rust than in most languages. Use them to model state machines, eliminate boolean flags, and make illegal states unrepresentable. Each variant can carry different data. Match expressions force you to handle all cases. This prevents entire categories of bugs at compile time.
📖 https://corrode.dev/blog/enums/
Avoid use module::* and minimize prelude usage. Glob imports make code harder to understand because you can't tell where names come from. They also cause name conflicts when modules add new exports. Explicit imports make dependencies clear and help with code navigation. The small amount of extra typing is worth it for maintainability.
📖 https://corrode.dev/blog/dont-use-preludes-and-globs/
Use the type system to enforce constraints at compile time instead of runtime. Create newtype wrappers that validate data in their constructors. Use const generics for array sizes. Encode state in types so invalid transitions don't compile. The more you can prove at compile time, the fewer runtime checks and tests you need.
📖 https://corrode.dev/blog/compile-time-invariants/
Design types so invalid combinations are impossible to construct. Instead of a struct with ssl: bool and ssl_cert: Option<PathBuf>, use an enum with Insecure and Ssl { cert: PathBuf } variants. This way the compiler prevents you from having SSL enabled with no certificate. Good type design eliminates whole categories of bugs.
📖 https://corrode.dev/blog/illegal-state/
Use this checklist to systematically review code. For each issue found, fetch the relevant blog article with webfetch to provide detailed context in your explanation.
as for narrowing type conversions instead of TryFromunwrap(), expect(), index[] without clear justification== instead of constant-time comparison_ catch-all#[must_use] on important return typesFrom implementations that can fail (should be TryFrom)Default trait (using ..Default::default())String, i32 instead of newtypes)let mut when value never changes)webfetch:
webfetch("https://corrode.dev/blog/simple/")
When you find code using as for casting:
as for narrowing conversions"webfetch("https://corrode.dev/blog/pitfalls-of-safe-rust/")TryFrom approachRemember: The blog articles contain detailed explanations, code examples, and rationale. Use them to provide educational, thorough feedback.