actionbook/rust-skills

What it solves

Rust Skills provides a structured approach to AI-assisted Rust development, moving beyond surface-level syntax fixes (like suggesting .clone() for ownership errors) to provide domain-correct architectural solutions. It prevents AI from giving generic answers by forcing it to reason through cognitive layers before recommending a fix.

How it works

The project implements a meta-cognition framework consisting of three layers:

1. Layer 1: Language Mechanics (How): Handles compiler rules and language features (e.g., ownership, concurrency).
2. Layer 2: Design Choices (What): Focuses on architectural patterns and ecosystem choices (e.g., crate selection, performance bottlenecks).
3. Layer 3: Domain Constraints (Why): Applies specific industry rules (e.g., FinTech audit trails or Embedded no_std constraints).

When a user asks a question, a hook layer triggers the rust-router to identify the entry layer and trace the logic upward or downward through these layers to reach a solution.

Who it’s for

Rust developers using AI coding agents (specifically Claude Code, but also compatible with Vercel AI and others) who want higher-quality, architecturally sound code suggestions rather than quick syntax patches.

Highlights

• Meta-Cognition Framework: A three-layer model that ensures AI reasons about domain and design before mechanics.
• Dynamic Skill Generation: Automatically generates AI skills based on the project's Cargo.toml dependencies.
• Real-time Data: Integrates with background agents to fetch the latest Rust versions and crate information.
• Broad Domain Support: Includes specialized constraints for FinTech, ML, Cloud-Native, IoT, Web, CLI, and Embedded development.
• Flexible Installation: Supports full plugin mode for Claude Code or a "skills-only" mode for other agents.