RSS
Tools

BeautifulRust: Agent Skills for writing, reviewing, and porting idiomatic Rust

BeautifulRust is a set of Agent Skills that teach coding agents to write, review, and port idiomatic Rust, with a tested clippy config to enforce the rules.

Pixel art banner of a crab on rocky tide pools at sunset for BeautifulRust agent skills
Image: morganlinton/BeautifulRust (MIT)

morganlinton/BeautifulRust gives a coding agent three skills for Rust: one for writing new code, one for reviewing existing code, and one for porting code from another language, plus a clippy configuration that enforces the rules a machine can check.

What it is

BeautifulRust is a set of standard Agent Skills, with folders under skills for rust-best-practices, rust-review, and rust-port, and a lints folder with a drop-in clippy configuration. The writing skill loads whenever an agent writes or changes Rust. The review skill looks at a file, a crate, or a diff and writes a prioritized report. The porting skill moves code from C, C++, Python, Go, or JavaScript to Rust while keeping the same behavior.

The guidance is distilled from five published sources, with a condensed digest of each bundled in the skill: the Google Rust Style Guide, the Rust API Guidelines, Effective Rust, Rust Design Patterns, and the Microsoft Pragmatic Rust Guidelines. Where the sources disagree, the skill picks a side and says why. Every code example compiles on stable Rust, and every lint was checked against current clippy.

In the README example, an agent asked for a small semver-parsing library without the skill produced working code with public fields, no Errors documentation, and unwrap in its doc examples. With the skill, the same agent kept the fields private, documented the failure modes, and wrote doc examples a reader can copy safely. Across three test tasks, the skill passed 29 of 30 checks against 24 of 30 without it.

Who it is for

It is for Rust developers who already use a coding agent and want that agent to follow the same style rules from one task to the next.

  • Claude Code, OpenAI Codex, GitHub Copilot, Cursor, and Gemini CLI, and any other agent that reads the Agent Skills format.
  • Writing a new crate or module, reviewing an existing file or pull request, or porting a module from another language into Rust.
  • Library and application code, which the skills treat differently: library code follows the API-design rules strictly, while application code favors concrete types and directness.
  • Bevy developers working with Bevy 0.20, for whom the skills read a separate Bevy reference first when a crate depends on bevy or bevy_ecs.

Quick facts

  • Platform: no operating system limit is stated in the README; it installs into the coding agents listed above.
  • Minimum versions: Node 22.20 or later for the npx skills installer, GitHub CLI 2.90 or later for the Copilot install, and rust-version 1.81 or later for two of the suppression lints.
  • License: MIT.
  • Cost: no price is listed in the README.
  • Language: Rust.
  • Latest release: v1.5.0, published on October 30, 2026, named v1.5.0: Bevy, and a benchmark across agents.
  • Stars: about 1,900 GitHub stars as of November 17, 2026.

One point needs care. The CHANGELOG and both plugin manifests already list version 1.6.0, dated November 8, 2026, with an MCP server and a Codex and ChatGPT plugin. That version is not a GitHub release: the newest entry on the Releases page I read was v1.5.0, so v1.5.0 is the latest release.

What stands out

  • The writing skill is organized so an agent can scan it while coding, covering mindset, a pre-handback checklist, ownership and borrowing, types and API design, errors and panics, idioms, concurrency and async, unsafe code, crates and dependencies, documentation, tests, and performance.
  • The review skill writes a report in three tiers, must fix, should fix, and consider. Every finding gives a file and line location, the rule it breaks, and the smallest fix, and each report ends with what the code already does well.
  • The porting skill ports the original tests first, designs a Rust API instead of transliterating the original, and writes a PORTING.md listing every deliberate difference. It maps NULL and None to Option, error codes and exceptions to an error enum, and out-parameters to return values.
  • The Bevy 0.20 reference leaves Bevy conventions alone, including system parameters taken by value, the Bevy prelude, and public fields on plain-data components. It also lists ECS bugs that compile cleanly, including two queries that can mutate the same component and panic with error B0001, deferred Commands, and an unwrapped single call.
  • The clippy configuration is a drop-in pair of files, clippy.toml and a lints block for Cargo.toml, and it is proven to fire: a check crate holds one deliberate violation per lint, and a script runs clippy on it and fails if a lint stops firing. It flags unwrap outside tests, unsafe blocks without a SAFETY comment, allow without a reason, bool parameters, and lossy as casts.

The October 2026 benchmark ran all eight eval cases on Claude Code and Codex CLI, with and without the skills, and graded the results blind. Claude Code moved from 72 of 85 to 80 of 85, and Codex CLI moved from 53 of 85 to 76 of 85. That was one run per setup.

Quick start

From the project's README, install all three skills; the review and port skills lean on the best-practices skill for style. The README is the install page, and the Releases page lists the released versions.

npx skills add morganlinton/BeautifulRust
claude plugin install beautiful-rust --marketplace morganlinton/BeautifulRust
codex plugin marketplace add morganlinton/BeautifulRust
codex plugin add beautiful-rust@beautiful-rust
gh skill install morganlinton/BeautifulRust rust-best-practices
gh skill install morganlinton/BeautifulRust rust-review
gh skill install morganlinton/BeautifulRust rust-port
gemini skills install https://github.com/morganlinton/BeautifulRust --path skills/rust-best-practices
gemini skills install https://github.com/morganlinton/BeautifulRust --path skills/rust-review
gemini skills install https://github.com/morganlinton/BeautifulRust --path skills/rust-port

For Claude Code, the plugin install above makes the skills appear as beautiful-rust:rust-best-practices, beautiful-rust:rust-review, and beautiful-rust:rust-port, though the README says you rarely need to call them, since they trigger on their own. Restart the agent if a skill does not show up right away. Claude Code does not read the shared .agents/skills location, so use its own directory or the plugin. To enforce the rules with clippy, copy lints/clippy.toml next to your Cargo.toml, paste the lints block from lints/lints.toml into it, then run:

cargo clippy --all-targets --all-features -- -D warnings

How it was built with Claude

Commits in the history I read carry Co-Authored-By trailers naming Claude. The tree I read has no CLAUDE.md or AGENTS.md file.

Try it with your AI. Paste this into your own coding assistant to try BeautifulRust on your own Rust code, filling in the square brackets.

Install BeautifulRust for [your coding agent, for example Claude Code or Codex] in [your
Rust project path], using only the official install command for that agent from the
BeautifulRust README, for example npx skills add morganlinton/BeautifulRust. Do not use
any other installer, script, or download source. Before you run anything, list every
command you plan to run, in order, and wait for me to confirm them. Do not install
anything from unknown sources, do not open or change any credential file, and do not
delete anything. After installing, use the rust-best-practices skill and the rust-review
skill on [the Rust file or crate you want checked, for example src/lib.rs]. Show me the
findings with the file and line, the rule each finding breaks, and the smallest fix, and
leave the code unchanged until I say which fixes to apply. If a step needs a Rust
toolchain, Node 22.20 or later, or GitHub CLI 2.90 or later, tell me first and wait for
confirmation.

Known limits

  • The skills collect no data, contain no analytics and no telemetry, and make no network requests, as the privacy policy states. Whatever you share with your agent is covered by that agent provider's privacy policy, not by this project.
  • Your agent may run local tools: when a Rust toolchain is installed, the skills ask it to run cargo fmt, cargo clippy, and cargo test in your project. Cargo can download crates your project depends on and runs your project's build scripts, as it always does.
  • The CHANGELOG entry for 1.6.0 describes an MCP server with three read-only tools, search_guidance, explain_lint, and get_crate_versions, which needs no account and no sign-in. The privacy policy says the server receives only the inputs of each tool call and stores nothing, while Vercel, which hosts the server, processes each request, including your IP address.
  • The benchmark was one run per configuration, graded by a Claude model, and the skills cost tokens: Claude used about 26 percent more tokens per task with the skills loaded, and Codex about 84 percent more input tokens. The skills also made Claude worse in one case, the Bevy feature task, where it scored 7 of 10 with the skills against 10 of 10 without them.
  • The Bevy reference covers Bevy 0.20, and it says to follow the version in Cargo.lock, not the reference, when they disagree. Bevy crates need four extra lint allows, listed in the Bevy reference.
  • Two suppression lints, allow_attributes and allow_attributes_without_reason, need rust-version 1.81 or later, because clippy switches them off for crates that cannot use expect. Lints marked lib are for libraries, and applications can drop them.
  • The eval cases are real model runs and are billed, as the README states.

Problems and feedback

Issues and pull requests are welcome, and the contributing guide says every rule needs a published source. The repository has issue templates for proposing a rule and for reporting a review false positive, and the privacy policy says questions go to GitHub issues.

Source: BeautifulRust repository, BeautifulRust README, BeautifulRust Releases, BeautifulRust CHANGELOG, BeautifulRust LICENSE, BeautifulRust PRIVACY.md, BeautifulRust CONTRIBUTING.md, BeautifulRust Issues, BeautifulRust cross-agent benchmark, October 2026, BeautifulRust Bevy reference, BeautifulRust lints.toml, BeautifulRust plugin.json, BeautifulRust rust-best-practices skill, BeautifulRust rust-review skill, BeautifulRust rust-port skill

This story first appeared in the claudefox daily briefing for November 17, 2026.