Execute qualquer Skill no Manus
com um clique

Execute qualquer Skill no Manus com um clique

mlx-optimizer

Estrelas0

Forks0

Atualizado12 de junho de 2026 às 11:33

Route Python-first MLX optimization work on Apple Silicon to focused audit, training, inference, Metal, or bridge workflows.

Instalação

Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.

Executar no Manus

Fonte

sealad886

sealad886/mlx-optimizer-plugin

Abrir repositório GitHub Ver repositórios do creator

Download

Executar no Manus

Ocupações relacionadasSOC

Baseado na classificação ocupacional SOC

Desenvolvedores de softwareInformática e Matemática·SOC 15-1252

SKILL.md

readonly

name	mlx-optimizer
description	Route Python-first MLX optimization work on Apple Silicon to focused audit, training, inference, Metal, or bridge workflows.

MLX Optimizer

Use this skill when the user asks Codex to optimize, audit, benchmark, profile, or explain MLX code on Apple Silicon.

First Actions

Identify the target repo or file.
Check for a repo-local .venv before any Python execution.
Inspect project structure before proposing new modules.
Treat static observations as candidate findings until verified.
Require concrete evidence for optimization claims.
For version-sensitive MLX API behavior, retrieve current authoritative MLX docs before implementation or recommendations.

Routing

Repo-wide or unknown performance issue: read ../mlx-performance-audit/SKILL.md.
Training loop issue: read ../mlx-training-optimizer/SKILL.md.
Inference or generation issue: read ../mlx-inference-optimizer/SKILL.md.
Metal capture, custom kernel, or C++ extension question: read ../mlx-metal-kernels/SKILL.md.
Swift/C/C++ or other language integration: read ../mlx-portability-bridges/SKILL.md.

Core References

Read only the files needed for the routed task:

../../references/mlx-core-concepts.md
../../references/eval-and-synchronization.md
../../references/compile-and-transforms.md
../../references/memory-and-dtypes.md
../../references/reporting-format.md

Non-Negotiables

Never install Python packages globally.
Preserve unrelated user changes.
Prefer existing repo abstractions.
Do not recommend automatic rewrites without active code-path evidence.
Include commands, outcomes, and residual risks in the final handoff.

Mais deste repositório

mesmo repositório

mlx-inference-optimizer

sealad886/mlx-optimizer-plugin

Optimize Python MLX inference and generation loops with warmup, batching, cache handling, synchronization, quantization, and memory checks.

2026-06-120

mlx-metal-kernels

sealad886/mlx-optimizer-plugin

Guide MLX Metal profiling, mx.fast escalation, custom Metal kernels, and C++ extensions when profiling proves kernel-level bottlenecks.

2026-06-120

mlx-performance-audit

sealad886/mlx-optimizer-plugin

Audit Python MLX repos for lazy-eval, synchronization, compile, dtype, memory, progress, and benchmark issues.

2026-06-120

mlx-portability-bridges

sealad886/mlx-optimizer-plugin

Advise on Python-first MLX integration with Swift, C, C++, and non-native language boundaries.

2026-06-120

mlx-training-optimizer

sealad886/mlx-optimizer-plugin

Optimize Python MLX training loops with value_and_grad, accumulation, checkpointing, dtype, validation cadence, memory telemetry, and progress reporting.

2026-06-120

name	mlx-optimizer
description	Route Python-first MLX optimization work on Apple Silicon to focused audit, training, inference, Metal, or bridge workflows.

MLX Optimizer

Use this skill when the user asks Codex to optimize, audit, benchmark, profile, or explain MLX code on Apple Silicon.

First Actions

Identify the target repo or file.
Check for a repo-local .venv before any Python execution.
Inspect project structure before proposing new modules.
Treat static observations as candidate findings until verified.
Require concrete evidence for optimization claims.
For version-sensitive MLX API behavior, retrieve current authoritative MLX docs before implementation or recommendations.

Routing

Repo-wide or unknown performance issue: read ../mlx-performance-audit/SKILL.md.
Training loop issue: read ../mlx-training-optimizer/SKILL.md.
Inference or generation issue: read ../mlx-inference-optimizer/SKILL.md.
Metal capture, custom kernel, or C++ extension question: read ../mlx-metal-kernels/SKILL.md.
Swift/C/C++ or other language integration: read ../mlx-portability-bridges/SKILL.md.

Core References

Read only the files needed for the routed task:

../../references/mlx-core-concepts.md
../../references/eval-and-synchronization.md
../../references/compile-and-transforms.md
../../references/memory-and-dtypes.md
../../references/reporting-format.md

Non-Negotiables

Never install Python packages globally.
Preserve unrelated user changes.
Prefer existing repo abstractions.
Do not recommend automatic rewrites without active code-path evidence.
Include commands, outcomes, and residual risks in the final handoff.