Applies five prompt patterns that reduce common bugs in LLM-generated Rust code.
npx skills add https://github.com/avbel/ai-skills --skill power-rust-avbel
Please help me install this Agent Skill. Skill: power-rust Source: https://github.com/avbel/ai-skills/tree/main/skills/power-rust Command: npx skills add https://github.com/avbel/ai-skills --skill power-rust-avbel
// cancel-safe: comments on async fns and // SAFETY: invariant lists on every unsafe block, creating a mechanical audit trail. - Lifetime call-site examples and trait design checkpoints: Demand example call sites before committing to non-trivial lifetime signatures, and require 2-3 candidate trait designs with object-safety and blanket-impl analysis before implementation. - Use Case: When asking a coding agent to write an HTTP handler, you specify "axum 0.7, tokio 1.35, sqlx 0.7 with postgres" and require cancel-safety annotations, catching a parking_lot::Mutex held across .await before it ships. ## Quick Start Ask your coding agent to write Rust using the power-rust prompt patterns, pinning crate versions and requiring cancel-safety and SAFETY annotations on the generated code.High-intent search queries and answers about installing and using this skill.
Pin exact crate versions and the async runtime in every prompt, require cancel-safety annotations on each async fn, and demand SAFETY comments on unsafe blocks. These prompt patterns measurably reduced error rates in a six-month benchmark of LLM Rust output.
State the exact major.minor version of every direct dependency, the async runtime and its version, and relevant feature flags, such as "axum 0.7, tokio 1.35, sqlx 0.7 with postgres". Without version anchors, models average across incompatible APIs.
Cancel-safety describes whether an async fn is safe to drop mid-future at an await point, and it has no signature marker. For example, tokio's AsyncReadExt::read is cancel-safe while read_exact is not, so each async fn should carry an explicit annotation naming the await point.
Yes, the Skill mandates clippy::pedantic at deny level for crates containing LLM-generated Rust, configured via [lints.clippy] in Cargo.toml. Pedantic lints catch residual mistakes like needless borrows, mis-sized integer casts, and redundant clones.
Run miri in nightly CI when code uses unsafe, FFI, or transmute, since it catches undefined behavior like uninitialized reads and pointer-provenance violations that clippy cannot see. cargo-careful is a lighter alternative when miri is too slow.
carayaFormat Rust code with rustfmt and apply Clippy linting.
sammcjAnalyze Rust workspace and Cargo configuration for safety and performance constraints.
aster-voidReview Rust code for ownership, error handling, and performance anti-patterns.
carayaAnalyze Rust ownership, borrowing, and lifetimes for safe, idiomatic code.