How to use the Rust Expert agent skill: install and trigger it

What is an agent skill and what does rust-expert make the agent do?

An agent skill is a folder containing a SKILL.md file of instructions that an AI agent can load when it is relevant to a task. The rust-expert skill's SKILL.md tells the agent to act as an expert Rust developer who writes code that is safe, performant and idiomatic.

In practice the agent follows guidance on core Rust ideas: who owns data, how long references live, how async tasks are spawned, selected and coordinated, how traits are designed, and when unsafe code is acceptable. The skill also points the agent toward common patterns such as the builder, newtype, RAII guard and typestate patterns.

• Prefer owned types at API boundaries and borrows inside function bodies
• Use thiserror for library errors and anyhow for application-level errors
• Use tokio::spawn, tokio::select! and tokio::sync::mpsc for concurrent work
• Avoid cloning just to satisfy the borrow checker, and avoid unwrap() in library code
• Document every unsafe block with a // SAFETY: comment

How do you install the rust-expert skill?

The skill is distributed as a ZIP file from this page. You can install it in one of two ways. For a personal setup, unzip it into ~/.claude/skills/ so the rust-expert folder sits there. For a single project, unzip it into the project's .claude/skills/ folder instead.

If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload the skill

How do you trigger the skill with a natural request?

You do not need a special command. Describe a Rust task in plain language and mention the topic, such as ownership, lifetimes, async, traits or unsafe. The agent can then load the guidance when it fits the request.

Being specific helps. Name the file or function, paste the code, and say whether you want it written, explained or reviewed.

• Review this Rust module for ownership problems and suggest borrows instead of clones
• Explain the lifetime on this function signature in simple terms
• Convert this struct to a builder pattern with a build() method that validates fields
• Check every unsafe block in this file for a // SAFETY: comment

What are good examples of using rust-expert?

The skill suits both writing and reviewing code, and it works well when you want an explanation alongside the change. The examples below are the kinds of tasks the guidance was written for.

• Refactor a function that clones a String to satisfy the borrow checker, so it uses a reference instead
• Design an error type with thiserror for a library crate, with clear #[error("...")] messages
• Review a tokio program for a MutexGuard held across an .await point
• Wrap raw String IDs in newtypes such as struct UserId(String) to stop mixing them up
• Teach a junior developer how Pin<Box<dyn Future>> is used when futures are stored in a struct

What does the skill need, and what about scripts and network use?

The skill lists no requirements. It does not run scripts and does not use the network, so it is a set of instructions only. Check the files before you install, since the package contains only SKILL.md.

The skill cannot check that generated code compiles or is correct. You should still build, test and review the output, and get expert human review for production-critical or unsafe systems code.

• Requires: nothing listed
• Runs scripts: no
• Uses network: no
• Only file included: SKILL.md

What tips help you get better results?

Give the agent your context: whether the crate is a library or an application, and which runtime you use. The guidance recommends different error handling for each, and it recommends tokio for async work, so say if you use something else.

• Say whether the code is a library or an application, since error-handling advice differs
• Ask the agent to explain its ownership and lifetime choices, not just the result
• Run cargo build and your tests on any change the agent proposes
• Ask for a // SAFETY: comment whenever unsafe code is involved