Rust Skills

265 Rust rules your AI coding agent can use to write better code. Current for Rust 1.96 (2024 edition).
Works with Claude Code, Cursor, Windsurf, Copilot, Codex, Aider, Zed, Amp, Cline, and pretty much any other agent that supports skills.
Why
Out of the box, coding agents write average Rust — they clone to dodge the borrow checker, .unwrap() everything, and reach for Box<dyn Trait> when impl Trait would do. These rules encode what expert Rust actually looks like: idiomatic, fast, and safe. Each rule is small and focused, so the agent pulls in only what's relevant to the code in front of it.
Install
npx add-skill leonardomso/rust-skills
That's it. The CLI figures out which agents you have and installs the skill to the right place.
How to use it
After installing, just ask your agent:
/rust-skills review this function
/rust-skills is my error handling idiomatic?
/rust-skills check for memory issues
/rust-skills is this unsafe block sound?
The agent loads the relevant rules and applies them to your code.
See it in action
Ask the agent to review a function like this:
// before
fn first_word(s: &String) -> String {
s.clone().split_whitespace().next().unwrap().to_string()
}
With these rules loaded, it knows to take &str instead of &String, drop the
needless clone() and allocation, and return an Option instead of panicking:
// after — applies own-slice-over-vec, own-borrow-over-clone, anti-unwrap-abuse
fn first_word(s: &str) -> Option<&str> {
s.split_whitespace().next()
}
What's in here
265 rules split into 26 categories:
| Category | Rules | What it covers |
|---|
| Ownership & Borrowing | 12 | When to borrow vs clone, Arc/Rc, lifetimes |
| Error Handling | 12 | thiserror for libs, anyhow for apps, the ? operator |
| Memory | 17 | SmallVec, arenas, avoiding allocations, mem::take, drop order |
| Unsafe Code | 7 | SAFETY: comments, Miri, MaybeUninit, 2024-edition unsafe |
| API Design | 17 | Builder pattern, newtypes, sealed traits, FromIterator |
| Async | 18 | Tokio patterns, channels, async fn in traits, cancel safety |
| Concurrency | 4 | rayon, scoped threads, atomic ordering, thread-locals |
| Optimization | 12 | LTO, inlining, PGO, SIMD |
| Numeric & Arithmetic | 5 | Overflow handling, as vs TryFrom, float compare, NonZero |
| Type Safety | 13 | Newtypes, parse don't validate, Deref, Display/Debug |
| Trait & Generics Design | 6 | dyn vs generic, associated types, blanket impls, object safety, orphan rule |
| Conversions | 3 | TryFrom, FromStr, AsMut |
| Const & Compile-Time | 4 | const fn, const vs static, const generics, const {} blocks |
| Serde | 8 | rename_all, default, flatten, enum tagging, validate-on-deserialize |
| Pattern Matching | 5 | let-else, matches!, if-let chains, exhaustive matches |
| Macros | 8 | macro_rules! hygiene, fragment specifiers, proc-macros with syn/quote |
| Closures | 5 | Fn/FnMut/FnOnce bounds, returning impl Fn, move & disjoint capture |
| Collections | 4 | HashMap/BTreeMap/IndexMap, Vec/VecDeque, sets, BinaryHeap |
| Naming | 16 | Following Rust API Guidelines |
| Testing | 15 | Proptest, mockall, criterion, loom, snapshot tests |
| Docs | 12 | Doc examples, intra-doc links, README/crate-doc unification |
| Observability | 7 | tracing over log, spans, structured fields, redacting secrets |
| Performance | 13 | Iterators, entry API, faster hashers, I/O buffering |
| Project Structure | 14 | Workspaces, module layout, features, MSRV |
| Linting | 13 | Clippy config, CI setup, unexpected_cfgs |
| Anti-patterns | 15 | Common mistakes and how to fix them |
Each rule has:
- Why it matters
- A bad code example
- A good code example
- Links to related rules and sources
How it works
The design is built for low token cost and easy auditing:
SKILL.md is a lightweight index — every rule listed as a one-line summary, grouped by category, with a link to its file. The agent reads this first.
rules/ holds one Markdown file per rule (<prefix>-<name>.md). The agent opens only the handful relevant to your code instead of loading all 218 — progressive disclosure keeps context small.
- Prefixes (
own-, err-, unsafe-, async-, …) map directly to categories, so an agent reviewing async code can pull just async-, conc-, and own- rules.
CLAUDE.md and AGENTS.md are symlinks to SKILL.md, so the same content works across agent conventions.
Manual install