claude-skill-rust

name	claude-skill-rust
description	Guide for Rust development including code style, testing, building, and quality checks using cargo tools. Apply when working with Rust code, Cargo.toml, or running cargo commands.
user-invocable	true
allowed-tools	Read, Grep, Bash

Rust Development Guide

Code Style

• Follow standard Rust conventions and idioms
• Use rustfmt for code formatting
• Configure lints in lib.rs (see Crate-Level Lint Configuration below)
• Prefer descriptive variable and function names

Module File Organization

Prefer the modern <module_name>.rs style over the legacy mod.rs style for module files.

Preferred (modern style):

src/
├── lib.rs
├── config.rs        # mod config
├── config/
│   └── parser.rs    # mod config::parser
├── network.rs       # mod network
└── network/
    ├── client.rs    # mod network::client
    └── server.rs    # mod network::server

Avoid (legacy style):

src/
├── lib.rs
├── config/
│   ├── mod.rs       # mod config
│   └── parser.rs    # mod config::parser
└── network/
    ├── mod.rs       # mod network
    ├── client.rs    # mod network::client
    └── server.rs    # mod network::server

Benefits of the modern style:

• File names directly indicate the module name (no ambiguous mod.rs files)
• Easier navigation in editors and file browsers
• Clear correspondence between module path and file path
• Supported since Rust 2018 edition

Module Visibility and Re-exports

Do not export internal modules with pub mod. Keep modules private (mod foo;) and publish only a curated public surface by re-exporting the specific items at the crate root with pub use. This keeps the module layout an implementation detail you can refactor freely, gives users one flat, stable import path (use my_crate::Thing;), and prevents internal helper types from leaking into the public API through the module namespace.

Preferred:

// lib.rs
mod packed;                       // private module
mod widget;                       // private module

pub use packed::{PackedArray, PackedVec};   // curated public surface
pub use widget::Widget;

#[cfg(feature = "alloc")]
pub use packed::PackedBitsVec;    // gate re-exports the same way as the items

Avoid:

pub mod packed;   // exposes the whole module namespace and its path structure
pub mod widget;   // as public API, including items you only meant to share
                  // internally

Guidance:

• Users should import from the crate root (use my_crate::PackedArray), never through a module path (use my_crate::packed::PackedArray).
• Gate a re-export with the same #[cfg(...)] as the item it re-exports.
• If an implementation-detail trait or type must be pub only because it appears in a public item's signature (Rust rejects a more-private type in a public interface, E0446), leave it pub but do not re-export it. Living in a private, non-re-exported module makes it unreachable to downstream crates even though the keyword is pub. Sealing it (a pub trait with a private supertrait) additionally prevents external implementations.
• The exception is a module that is intended as a namespaced public API grouping (e.g. prelude); make that pub deliberately, not by default.

Crate-Level Lint Configuration

Define lint rules at the top of lib.rs (or main.rs for binaries) rather than via command-line flags. This ensures consistent enforcement and documents project standards.

#![deny(unsafe_code)]
#![cfg_attr(all(not(debug_assertions), not(test)), deny(clippy::all))]
#![cfg_attr(all(not(debug_assertions), not(test)), deny(clippy::pedantic))]
#![cfg_attr(all(not(debug_assertions), not(test)), deny(missing_docs))]
// Allow some pedantic lints that are too strict for this project
#![allow(clippy::module_name_repetitions)]
#![allow(clippy::must_use_candidate)]
#![allow(clippy::missing_errors_doc)]
#![allow(clippy::missing_panics_doc)]
#![allow(clippy::enum_variant_names)]
// Until 1.0.0, allow dead code and unused dependency warnings
#![allow(dead_code)]
#![allow(unused_crate_dependencies)]

Key principles:

• Deny unsafe_code unless explicitly required
• Use cfg_attr to deny lints only in release builds (not debug or test)
• Allow specific pedantic lints that conflict with project conventions
• Document why each allow is necessary

Important: These are good defaults for new crates. Do not override existing lint configurations - if a crate already specifies deny or allow for a rule, respect that choice.

Error Handling

Avoid .unwrap() and .expect() - these cause panics and should not be used in production code. Instead:

• Use Result<T, E> return types with the ? operator for error propagation
• Prefer early returns for error conditions
• Handle errors explicitly at appropriate boundaries

Error Type Crates

Note: The following recommendations apply to std crates only. For no_std embedded targets, use custom error enums or thiserror with default-features = false.

• Libraries: Use thiserror to define custom error types with derive macros
• Applications: Use anyhow for convenient error handling with context

Example library error type with thiserror:

use thiserror::Error;

/// Errors that can occur during data processing.
#[derive(Debug, Error)]
pub enum ProcessError {
    /// The input data was empty.
    #[error("input data cannot be empty")]
    EmptyInput,

    /// Failed to parse the input data.
    #[error("failed to parse input: {0}")]
    ParseFailure(String),

    /// An I/O error occurred.
    #[error("I/O error: {0}")]
    Io(#[from] std::io::Error),
}

Example application error handling with anyhow:

use anyhow::{Context, Result};

fn load_config(path: &Path) -> Result<Config> {
    let contents = std::fs::read_to_string(path)
        .context("failed to read config file")?;

    let config: Config = serde_json::from_str(&contents)
        .context("failed to parse config JSON")?;

    Ok(config)
}

Logging

Note: The following recommendations apply to std crates only. For no_std embedded targets, use defmt or platform-specific logging mechanisms.

Avoid println! and eprintln! outside of main application entry points. Use the log crate for structured logging instead.

use log::{debug, info, warn, error};

fn process_data(data: &[u8]) -> Result<(), ProcessError> {
    debug!("Processing {} bytes", data.len());

    if data.is_empty() {
        warn!("Received empty data buffer");
        return Err(ProcessError::EmptyInput);
    }

    info!("Data processed successfully");
    Ok(())
}

Log levels:

• error! - Unrecoverable errors or failures
• warn! - Unexpected conditions that are handled
• info! - Significant events in normal operation
• debug! - Detailed information for debugging
• trace! - Very detailed tracing information

Applications should initialise a log implementation (e.g., env_logger, tracing-subscriber) in main().

Whitespace and Formatting

Maintain blank lines for readability:

• Between module-level items (functions, structs, enums, traits, constants, impl blocks)
• Between struct and enum members when they have doc comments
• Between constant declarations
• After use statements before the first item
• After code blocks (loops, conditionals, match arms) before subsequent statements

Struct Field Formatting

When struct or enum fields have documentation comments, add a blank line before each doc comment to visually separate the fields:

pub struct Handle<T> {
    /// Index into the arena's resource vector.
    index: usize,

    /// Generation counter to detect stale handles after resource reuse.
    generation: u32,

    /// Type marker to prevent mixing handles of different resource types.
    _marker: PhantomData<T>,
}

For fields without documentation, blank lines are optional but may still aid readability for complex types.

Documentation

All items require rustdoc documentation comments.

For documenting the following item (functions, types, constants, fields):

/// Documentation for the item below.

For documenting the enclosing item (modules, crate root):

//! Documentation for the containing module.

Items requiring documentation include:

• Functions and methods
• Types (structs, enums, type aliases)
• Traits and trait implementations
• Constants and statics
• Modules
• Test functions
• Struct and enum fields

Function Documentation

Function documentation must include:

• Brief description of purpose
• # Arguments section documenting each parameter
• # Returns section documenting the return value
• # Errors section if the function returns Result
• # Panics section if the function can panic

Example:

/// Calculates the checksum for the given data buffer.
///
/// # Arguments
///
/// * `data` - The byte slice to calculate the checksum for.
/// * `seed` - Initial seed value for the checksum algorithm.
///
/// # Returns
///
/// The computed 32-bit checksum value.
///
/// # Errors
///
/// Returns `ChecksumError::EmptyBuffer` if `data` is empty.
fn calculate_checksum(data: &[u8], seed: u32) -> Result<u32, ChecksumError> {
    // ...
}

Constant Documentation

Constant documentation must include:

• Description of the constant's purpose
• Reference to specification, document, or section where applicable

Example:

/// Maximum transmission unit size in bytes.
///
/// Per RFC 894, Section 3 - Ethernet frames have a maximum payload of 1500 bytes.
const MTU_SIZE: usize = 1500;

Testing

• Write unit tests in the same file as the code being tested
• Use integration tests in the tests/ directory for end-to-end functionality
• Run tests with cargo test
• Aim for meaningful test coverage of core functionality
• Document test functions with their purpose and what they verify

Build Verification

After completing changes to Rust code, run the following checks in order:

1. Format Code: cargo fmt
2. Format Check: cargo fmt --check
3. Lint Check: cargo clippy -- -D warnings
4. Tests: cargo test
5. Release Build: cargo build --release
6. Documentation: cargo doc --no-deps --document-private-items
7. License Check: cargo deny check
8. Security Audit: cargo audit

If any check fails, fix the issues before proceeding.

Security Tools

Projects should use these security tools:

• cargo-deny: License and advisory checker. Install with cargo install cargo-deny. Configuration in deny.toml enforces permissive licenses (MIT, Apache-2.0, BSD) and denies copyleft licenses (GPL, LGPL, AGPL).

• cargo-audit: Security vulnerability scanner. Install with cargo install cargo-audit. Scans dependencies against the RustSec Advisory Database.

Dependencies

When adding dependencies:

• Choose well-maintained crates widely used in the Rust ecosystem
• For workspaces, add shared dependencies to [workspace.dependencies] in root Cargo.toml
• Justify each dependency with a clear use case
• Disable default features and explicitly enable only required features

Minimising Dependencies

Always add dependencies with default-features = false and explicitly specify the features you need. This reduces compile times and binary size by avoiding unnecessary transitive dependencies.

# Prefer this:
serde = { version = "1.0", default-features = false, features = ["derive"] }

# Avoid this:
serde = "1.0"

When adding a new dependency:

1. Check the crate's documentation for available features
2. Identify the minimum set of features required for your use case
3. Add with default-features = false
4. Explicitly list only the features you need