Agent-Skills.md

Agent Skills: Rust Development

Idiomatic Rust development. Use when writing Rust code, Cargo crates/workspaces,

UncategorizedID: alexei-led/claude-code-config/writing-rust

Repository

alexei-led/claude-code-config
alexei-ledLicense: MIT
355

Install this agent skill to your local

pnpm dlx add-skill https://github.com/alexei-led/cc-thingz/tree/HEAD/src/skills/writing-rust

Skill Files

Browse the full folder contents for writing-rust.

Download Skill

Loading file tree…

src/skills/writing-rust/SKILL.md

Skill Metadata

Name
writing-rust
Description
Idiomatic Rust development. Use when writing Rust code, Cargo crates/workspaces,

Rust Development

Covers crates, workspaces, and Rust CLIs. Check Cargo.toml, rust-toolchain.toml, and CI before using newer syntax or APIs. Project conventions win over these defaults.

Version and Features

  • Stay within package.rust-version (MSRV) and the pinned toolchain unless the task is an upgrade.
  • Use Edition 2024 syntax only when the crate sets edition = "2024".
  • Keep features additive. Do not assume --all-features compiles when the crate documents mutually exclusive features.

Defaults

  • Stdlib and existing crates first. thiserror for library error enums and anyhow for application glue only when the crate already uses them or boilerplate justifies it.
  • Newtypes for IDs, tokens, and validated values; enums for states. Private first, then pub(crate), then pub.
  • Implement From, TryFrom, AsRef, FromStr, and Display before ad hoc conversion methods.
  • Parsing and validation return Result with a typed error. unwrap/expect in production code only with a named local invariant.
  • When the borrow checker objects, fix the data flow before reaching for clone(). Use lifetimes in APIs only for real zero-copy needs.
  • Async only where the runtime or I/O boundary requires it. Never hold a blocking mutex guard across .await. Bound channels, queues, and retries.
  • Keep secrets out of Debug output, logs, errors, and snapshots.
  • unsafe: small blocks behind a safe API, a // SAFETY: comment stating the invariant, and focused tests or configured Miri.

Checks

Use the project's configured Cargo commands first. While editing, run focused checks (cargo check -p <crate>, cargo test -p <crate> <name>); before finishing, run the full gates:

cargo fmt --check
cargo clippy --all-targets -- -D warnings
cargo test --all-targets   # or cargo nextest run, when the project uses nextest
cargo test --doc           # --all-targets and nextest both skip doctests

Test behavior through the public API with valid, invalid, and boundary cases, not private helpers. Never run cargo clean as a routine fix; it throws away the incremental cache.

References

  • testing.md: read when adding or reshaping tests, or when the test loop is slow.
  • linting.md: read when changing rustfmt, Clippy, or workspace/feature check flags.

Done when the relevant build/test/lint checks pass on what you changed, or you name each check that did not run and why.