Use when starting any Rust task to route to the correct rust-* skill, and when finishing any Rust task to run the cross-skill quality checklist (clippy, rustfmt, MSRV, edition idioms, error handling, async hygiene, doc coverage). Prevents loading the wrong skill, missing a relevant skill, and shipping Rust code that fails clippy or ignores the project MSRV. Covers: a routing table mapping user-prompt patterns to rust-* skills (ownership question to rust-syntax-ownership, lifetime error to rust-errors-lifetimes, async runtime question to rust-impl-async-tokio, etc.), and a cross-skill quality checklist every Rust task should pass before completion. Keywords: "which Rust skill", "Rust task routing", "Rust code review checklist", "rust quality gate", clippy, rustfmt, MSRV, "Rust best practice check", "Rust code quality", "before I ship Rust", orchestrator, "Rust skill index", "what skill for", routing, "Rust checklist".
インストール
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
Use when starting any Rust task to route to the correct rust-* skill, and when finishing any Rust task to run the cross-skill quality checklist (clippy, rustfmt, MSRV, edition idioms, error handling, async hygiene, doc coverage). Prevents loading the wrong skill, missing a relevant skill, and shipping Rust code that fails clippy or ignores the project MSRV. Covers: a routing table mapping user-prompt patterns to rust-* skills (ownership question to rust-syntax-ownership, lifetime error to rust-errors-lifetimes, async runtime question to rust-impl-async-tokio, etc.), and a cross-skill quality checklist every Rust task should pass before completion. Keywords: "which Rust skill", "Rust task routing", "Rust code review checklist", "rust quality gate", clippy, rustfmt, MSRV, "Rust best practice check", "Rust code quality", "before I ship Rust", orchestrator, "Rust skill index", "what skill for", routing, "Rust checklist".
license
MIT
compatibility
Designed for Claude Code. Requires Rust 1.85+, edition 2024.
metadata
{"author":"OpenAEC-Foundation","version":"1.0"}
Rust Agents Orchestrator
This is the routing brain of the Rust skill package. It does two jobs:
Route: at the START of a Rust task, map the user prompt to the exact rust-* skill (or skills) to load.
Gate: at the END of a Rust task, run the cross-skill quality checklist before declaring the work done.
ALWAYS run step 1 before writing Rust code. ALWAYS run step 2 before reporting a Rust task complete.
This skill routes and checklists. It does NOT do deep lint review: that is rust-agents-code-reviewer. It does NOT iterate on compiler errors: that is rust-agents-compile-fix. Treating these three as interchangeable is an anti-pattern.
ALWAYS pick the most specific skill. If a prompt contains a rustc error code (E0xxx) or the word "error"/"fails to compile", route to an errors-* skill, NOT a syntax-* skill.
Ownership, borrowing, lifetimes
User prompt signal
Skill
"move", "value used after move", "who owns this"
[[rust-syntax-ownership]]
E0382, "borrow of moved value"
[[rust-errors-borrow-checker]]
"& vs &mut", "borrow", "two-phase borrow", "RefCell"
"review my Rust code", "is this idiomatic", deep clippy lint review
[[rust-agents-code-reviewer]]
"fix this compile error", iterate on rustc output, follow --explain suggestions
[[rust-agents-compile-fix]]
"which skill", "route this task", "run the quality checklist"
this skill
Decision Tree: Ambiguous Prompt to Skill
Rust task arrives
├── Contains an E0xxx code or "does not compile"? ......... ROUTE to errors-*
│ ├── E0382 / E0502 / E0596 / E0500 ............. rust-errors-borrow-checker
│ ├── E0106 / E0623 / E0495 / E0700 ............. rust-errors-lifetimes
│ ├── E0277 / E0599 / E0220 ..................... rust-errors-trait-bounds
│ ├── linker / native lib / version clash ....... rust-errors-build-link
│ ├── async `!Send` / Pin across .await ......... rust-errors-async
│ └── panic / OOB / abort at runtime ............ rust-errors-runtime
├── "How do I build / set up X"? ...................... ROUTE to impl-*
├── "Why does Rust work like X" (concept)? ............ ROUTE to core-*
├── "What is the syntax for X" (language form)? ....... ROUTE to syntax-*
├── "Review / is this idiomatic"? ..................... rust-agents-code-reviewer
└── Multiple skills plausible? ........................ load the NARROWEST
that names the exact construct, then load its core-* concept skill
only if the user needs the mental model, not just the syntax.
ALWAYS prefer the narrowest match. NEVER load a core-* overview when the user asked a concrete syntax-* or impl-* question. Load 1 to 3 skills, not the whole package.
Cross-Skill Quality Checklist
Run this BEFORE declaring ANY Rust task complete. Every item is a hard gate.
#
Check
Command / inspection
Pass criterion
a
Edition + MSRV
read edition and rust-version in Cargo.toml
code uses only features available at the stated MSRV; edition is 2024 unless the project pins older
b
Clippy
cargo clippy --all-targets -- -D warnings
no warnings; correctness + suspicious + perf clean
c
Formatting
cargo fmt --all --check
exits 0, no diff
d
No panics on fallible paths
grep .unwrap() / .expect() in non-test code
every remaining one is on a genuinely infallible path or has a justifying comment
e
Async hygiene
inspect .await points
no blocking call on the async executor; no !Send value held across .await in a spawned task
f
Doc coverage
cargo doc + inspect public items
every public item has a /// doc; add a doctest where it clarifies usage
g
Unsafe is justified
grep unsafe blocks
every unsafe block has a // SAFETY: comment stating the invariant
h
Error type contract
inspect library error types
error types implement std::error::Error + Display; a library does NOT leak anyhow::Error in its public API
i
Tests
cargo test --all-targets and cargo test --doc
all unit, integration, and doc tests pass
If any item fails, fix it and re-run the failing check. NEVER report a Rust task done with an unchecked item.
See references/methods.md for the full command set and CI wiring, references/examples.md for worked routing and checklist runs, references/anti-patterns.md for routing and gating mistakes.
Quick Reference
Routing happens FIRST, the checklist happens LAST. Never skip either.
Error code present means an errors-* skill, never a syntax-* skill.
cargo build passing is NOT the quality gate. cargo clippy plus cargo fmt --check plus cargo test is.
MSRV is set by rust-version in Cargo.toml. A new dependency must not raise it silently.
This skill routes and gates. Deep review is rust-agents-code-reviewer. Error iteration is rust-agents-compile-fix.
