Skip to content

AI Usage Recipes

These recipes give assistants a small, explicit RustUse context path before they answer. They are intentionally short so they can be pasted into chat tools, coding assistants, or local agents.

Use them as starting points, then verify any generated Rust code against RustUse Rustdocs and cargo test.

Use https://rustuse.org/llms.txt to understand the RustUse project structure before answering.

Use this when the assistant needs a quick map of facades, docs routes, and Rustdoc patterns.

Use https://rustuse.org/facades/use-math/llms.txt for focused context on the RustUse math facade.

Use this when the task is scoped to one facade and does not need the full RustUse catalog.

Current live route:

Use https://rustuse.org/facades/use-math/llms.txt for focused context on use-math, then verify use-prime examples against RustUse Rustdocs.

Planned crate-specific route:

Use https://rustuse.org/crates/use-prime/llms.txt before suggesting examples for use-prime.

Use this shape once crate-level LLM text files are generated. Until then, start from the owning facade file.

Before suggesting a new RustUse crate, check the RustUse catalog and avoid names that already exist.

For crate naming work, also check whether a proposed hyphenated or underscored name would collide with an existing crates.io package identity.

Use the RustUse conventions from the LLM text files and generate a GitHub Copilot Plan Mode prompt.

Use this when you want an assistant to produce a plan that respects RustUse facade boundaries, facade crates, focused child crates, docs, tests, and release-readiness checks.

Use the RustUse catalog, avoid existing names, keep the facade crate thin, and propose focused child crates that provide primitive utilities.

Good facade plans should include a narrow domain boundary, proposed child crates, naming rationale, non-goals, documentation shape, and validation steps.

After suggesting RustUse code, list the Rustdocs route and cargo command I should use to verify it.

This keeps the assistant’s answer tied to public docs and local Rust validation instead of treating generated output as final.