skills

Profile a model running on MAX to find where it spends time and whether the GPU is saturated. Use when the user asks to "profile my model," "where is my model spending time," "why is inference slow," "is my GPU being utilized," "how much GPU am I using," "get a kernel breakdown," "capture an nsys/rocprof/ncu trace of max serve," or wants to measure MAX inference performance. Works for any model MAX can run — built-in architectures and custom ones loaded with --custom-architectures — from a pip or pixi install (max generate, max serve, or a Python script) on NVIDIA or AMD GPUs. Decide cheapest-first: a GPU utilization check, then a kernel breakdown, then a single-kernel deep dive only when one kernel dominates.

Use when importing a new model architecture into MAX from a Hugging Face model ID. Triggers on: "import a model into MAX", "add model to MAX", "bring up <HF model> in MAX". Workflow: inspect Hugging Face config and modeling code, scaffold from a similar MAX architecture, implement each graph layer to match HF, serve, then verify against the Hugging Face reference. When the server runs but output is wrong (gibberish, greedy mismatch, coherent-then-diverges), load debug-model for the divergence hunt instead of scalar-tap iteration.

Debug silent corruption when a MAX model loads, compiles, serves, and generates tokens but output disagrees with a reference implementation. Use whenever parity debugging stalls on scalar taps, the model returns gibberish or wrong greedy tokens, logit cosine is high but argmax differs, or generation is coherent then diverges — during an architecture port, a quantization bring-up, a multi-GPU conversion, or after a MAX upgrade. Triggers on "parity failure", "silent corruption", "logits match but tokens diverge", "top-1 mismatch", "greedy divergence", and "model serves but generates garbage". Not for crashes on load or pre-serve scaffolding (use import-model). Mandates reference-vs-MAX tensor-dump comparators first, verify fixes numerically before recompiling, and serve-vs-pipeline bisect when dumps match but text diverges.

Help to write Mojo code using current syntax and conventions. Always use this skill when writing any Mojo code, including when other Mojo-specific skills (e.g., mojo-gpu-fundamentals) also apply. Use when writing Mojo code, translating projects to Mojo, or otherwise generating Mojo. Use this skill to overcome misconceptions with how Mojo is written.

Aids in writing Mojo code that interoperates with Python using current syntax and conventions. Use this skill in addition to mojo-syntax when writing Mojo code that interacts with Python, calls Python libraries from Mojo, or exposes Mojo types/functions to Python. Also use when the user wants to build Python extension modules in Mojo, wrap Mojo structs for Python consumption, or convert between Python and Mojo types.

The basics of how to program GPUs using Mojo. Use this skill in addition to mojo-syntax when writing Mojo code that targets GPUs or other accelerators. Use targeting code to NVIDIA, AMD, Apple silicon GPUs, or others. Use this skill to overcome misconceptions about how Mojo GPU code is written.

Creates a new Mojo or MAX project. Use when wanting to start a new Mojo or MAX project, initializing the Pixi or UV environment to use Mojo or MAX, or when the user wants to begin a new Mojo or MAX project from scratch.