Structured skill packs for AI agents (Claude Code, Codex, Cursor) to operate the Horizon AI toolchain — HBDK compilation, HMCT/Plugin quantization, UCP on-board inference, and LLM compression.
Skill Claude CodeCodex
An index of instructions for HBDK4, a toolchain that converts machine-learning models into files runnable on Horizon hardware. It covers importing, serializing, quantizing, modifying, compiling, inspecting, and validating models.
Skill Claude CodeCodex
A configuration-driven process for compiling ONNX or BC machine-learning models for J6 HBDK hardware. It creates a YAML configuration, waits for approval by default, then compiles and checks the output.
Skill Claude CodeCodex
A routing workflow for the HMCT model-conversion toolchain. It handles model quantization, quick conversion checks, accuracy tuning, and targeted debugging.
Skill Claude CodeCodex
A workflow for improving the numerical accuracy of quantized HMCT models. Quantization uses lower-precision numbers to make models smaller or faster, and cosine similarity compares their outputs with the original model.
Skill Claude CodeCodex
A routing skill for OpenExplorer and Horizon tools used to quantize, compile, deploy, and evaluate machine-learning models on Horizon hardware. It selects the relevant sub-skill and defines rules for complete deployment workflows.
Skill Claude CodeCodex
A setup skill that detects the hardware board used for board-side software work and records its connection and platform details in the project.
Skill Claude CodeCodex needs its repo
A setup checker for the OE-LLM package environment used by tools for large language model compression, compilation, quantization, and device inference.
Skill Claude CodeCodex needs its repo
A local installer for OE-LLM, a package of language-model components used with the OpenExplorer toolchain. It creates a Python virtual environment and installs compatible package files when the user approves.
Skill Claude CodeCodex
An environment checker for the OpenExplorer toolchain, a collection of tools for quantizing, compiling, deploying, and evaluating machine-learning models. It finds the package, records its version, checks the local setup, and writes the results to an environment file.
Skill Claude CodeCodex
A skill for installing OE package components locally in a Python virtual environment, an isolated Python setup, after OE package detection. It checks Python, CUDA, and PyTorch versions, selects matching wheel files, installs them, and sets local execution mode when the user agrees.
Skill Claude CodeCodex
A performance-analysis workflow for checking ONNX, HBIR, and HBM machine-learning models on Horizon Robotics J5 and J6 hardware. It examines measurements such as delay, memory bandwidth, and BPU use.
Skill Claude CodeCodex
Integration guidance for the OpenExplorer command-line toolchain, which converts and prepares machine-learning models for supported hardware. It covers configuration, compilation, quantization, accuracy checks, performance analysis, and deployment.
Skill Claude CodeCodex
Instructions for running many machine-learning model quantization experiments and comparing their results. Quantization reduces the numerical precision used by a model, which can make it smaller or faster but may affect accuracy.
Skill Claude CodeCodex
A workflow for running LightCompress model-quantization experiments and producing accuracy reports. Quantization reduces the numerical precision used by a machine-learning model to explore smaller or faster versions.
Skill Claude CodeCodex
A development workflow for adding support for a new language or vision-language model to the llmcompression framework. It compares the model with an existing implementation or with the Transformers source code before integration.
Skill Claude CodeCodex needs its repo
Um guia para as operações diárias da ferramenta llm_compression, usada para calibrar, avaliar, compilar e executar modelos de linguagem em placas. Um modelo de linguagem é um sistema de aprendizagem automática que processa e gera texto.
Skill Claude CodeCodex
An orchestration guide for adapting a floating-point PyTorch model to Horizon Robotics' quantization toolkit, horizon_plugin_pytorch. Quantization changes a model so it can use lower-precision numbers, and QAT means quantization-aware training.
Skill Claude CodeCodex
A guide for adding dynamic-block markers when adapting PyTorch models for Horizon Robotics' quantization toolkit. Dynamic control flow means loops or branches whose execution changes with the input, random values, or external state.
Skill Claude CodeCodex
A code-editing guide for adding deployment-boundary markers to floating-point PyTorch models. It places separate QuantStub and DeQuantStub modules around the part of the model that will be quantized by Horizon's tools.
Skill Claude CodeCodex
A code-editing guide for preparing floating-point PyTorch models for Horizon's quantization-aware training tools. It adds the required prepare call and uses a fixed configuration for double-int8 operations.
Skill Claude CodeCodex
A coding guide for choosing the fake-quantization state of a model in Horizon's PyTorch quantization workflow. Fake quantization simulates reduced-precision calculations so a model can be prepared for quantized deployment.
Skill Claude CodeCodex
A coding guide for setting the target BPU hardware version in Horizon's PyTorch quantization and deployment workflows.
Skill Claude CodeCodex
A troubleshooting procedure for finding where a Horizon QAT PyTorch model loses accuracy between training and hardware deployment. It compares the model and outputs at stages from the trained file through conversion, compilation, and the device.
Skill Claude CodeCodex
A procedure for creating a separate script that exports a Horizon QAT PyTorch model into HBIR, an intermediate format used for later compilation and deployment. QAT means quantization-aware training, where a model is trained to account for reduced-precision calculations.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: