SuperAPI AI Directory

rust-mega-eng

Coding & development

View repository

Updated 10-15 17:27

About this skill

Rust Mega Eng

Use this skill as the explicit orchestration layer for broad Rust architecture and product engineering work. For ordinary local Rust edits, use rust-expert or a specialist skill instead.

Invocation Rule

Do not invoke this skill implicitly for normal Rust questions, compiler errors, single-crate fixes, broad architecture wording, or isolated CLI/TUI/Tauri/service work. Invoke only when the user explicitly names rust-mega-eng for broad, cross-cutting Rust architecture and execution planning.

Operating Model

  1. Map the product surface: libraries, binaries, services, desktop apps, terminal UI, deployment targets, public APIs, and release obligations.
  2. Identify canonical ownership: crate boundaries, app/service boundaries, shared domain types, generated code, test fixtures, config, docs, and CI gates.
  3. Choose leverage first: mature crates, official frameworks, platform conventions, and repo-native workflows before custom infrastructure.
  4. Split delivery into reviewable lanes: architecture/design, code movement, behavior changes, test hardening, release/distribution, and docs.
  5. Keep a decision ledger for major choices, with explicit tradeoffs and verification evidence.

Reference Map

  • references/orchestration.md for workspace strategy, crate portfolio defaults, decision scoring, fanout planning, release engineering, and anti-patterns.

Use the specialist skills for detailed implementation:

  • rust-cli-clap for CLI and clap command surfaces.
  • rust-tui-ratatui for terminal UI architecture.
  • rust-tauri-apps for Tauri v2 Rust app backends and IPC.
  • rust-web-services for Axum/Tokio/Tower services.
  • rust-expert for core Rust ownership, async, errors, crates, testing, performance, and security.

Verification

For broad Rust work, define a validation ladder early:

cargo fmt --all --check
cargo clippy --all-targets --all-features -- -D warnings
cargo test --all-targets --all-features
cargo nextest run --all-targets --all-features
cargo deny check
cargo audit
cargo semver-checks check-release

Run only gates that are relevant and available, but report any skipped gate with the reason.