Mojo (Flow Tools)

<workflow>

🚀 Official Modular Skills (Highly Recommended)

For comprehensive support for modern Mojo syntax, project initialization, and GPU programming, we highly recommend installing the official Modular agent skills:

• mojo-syntax: Overcomes agent misconceptions and ensures correct modern syntax.
• new-modular-project: Wizard for initializing Mojo/MAX projects with Pixi and UV.
• mojo-python-interop: Expert guidance for zero-copy Python interaction.
• mojo-gpu-fundamentals: Patterns for high-performance accelerator programming.

Installation:

npx skills add modular/skills

Supplemental Patterns

The patterns below focus on project integration and build hooks.

Hatch-Mojo Build Hook

The hatch-mojo plugin allows seamless compilation of Mojo source files during the standard Python build process.

Key Configuration (pyproject.toml):

[tool.hatch.build.targets.wheel.hooks.mojo]
# Configuration for mojo compilation

</workflow> <guardrails> ## Guardrails

• Explicitly define memory ownership -- Use owned, borrowed, and inout to manage data flow and avoid unnecessary copies.
• Use SIMD for performance-critical logic -- Mojo excels at vectorization; always consider SIMD when processing large arrays.
• Verify data alignment -- Ensure pointers are aligned for the target architecture, especially when using SIMD operations. </guardrails>

<validation> ## Validation Checkpoint

• Explicit ownership markers are correctly applied to arguments
• SIMD vectorization is implemented where applicable
• Memory safety is verified through ownership and borrowing checks </validation>

<example> ## Hatch-Mojo Build Hook Example

[tool.hatch.build.targets.wheel.hooks.mojo]
dependencies = ["hatch-mojo"]
path = "src/my_extension.mojo"

</example>