amd

The non-obvious invariants of the proactive nightly Claude audit workflow (.github/workflows/claude-nightly-audit.yml): the day-of-week mode selector, the two-key dedup scheme (cluster_key per defect, dedup_key per location) that keeps one defect from…

Use when integrating the @amd-gaia/gaia npm package — running GAIA's flagship agent, or embedding its local sidecar into a Node, TypeScript, or Electron app. Covers the one-command path, the SHA-256 integrity gate, platform coverage, starting the sidecar, the…

Cut a GAIA release end-to-end: draft notes, open release PR, run pre-tag verification, push the tag, monitor the publish pipeline, and produce the Discord announcement. Use when the user asks to 'cut a release', 'release vX.Y.Z', 'tag a release', or 'publish…

Use when integrating the @amd-gaia/agent-email npm package — embedding the GAIA email agent (a local triage/draft/send sidecar) into a Node, TypeScript, or Electron app. Covers install, spawning the sidecar, calling the typed client, prerequisites, and the…

Analyze your own local Claude Code session transcripts to find what you actually use the agent for, how effective it is, where it fails, and what it costs. Use when asked to mine Claude Code sessions, extract use-cases or workflows from session history,…

Use when testing or validating the GAIA TUI (tui/) by actually running it — driving screens, sending queries to an agent, verifying UX. Covers the loopback control API, waiting on state instead of sleeping, and the capture mistakes that produce false bug…

Cut or wire a frozen-binary + npm sidecar release for a GAIA agent (the email-agent CI pipeline: freeze -> Agent Hub Worker /publish -> npm OIDC). Use when releasing or onboarding a sidecar agent under hub/agents/, or authoring a release_agent_<id>.yml. For…

Test the flagship GAIA agent end-to-end through the Go TUI — launch it, drive it without colliding with other agents, run the capability ladder, verify skills really call their tools instead of fabricating, and check the permission gate. Use when validating…

Collect and normalize environment facts (OS, Python, GPU, CUDA/ROCm, container state) before Quark installation or PTQ planning. Use this skill whenever a downstream skill needs confirmed hardware and toolchain facts, when the user mentions their setup, when…

Validate workspace shape, model paths, output directories, and repo structure before downstream Quark skills proceed. Use this skill whenever a skill needs confirmed file paths, when the user provides a model path or output directory, when you need to…

Diagnose failed Quark ONNX installation, calibration, quantization, custom-op compilation, or export attempts. Use when the user reports an error, stack trace, invalid artifact, missing dependency, ORT execution-provider mismatch, silent CPU fallback, OOM…

Install or verify the correct ONNX Runtime build (and the `onnx` package) for a user's accelerator backend before Quark's ONNX-to-ONNX flow. Use when the user needs ONNX Runtime set up, reports onnxruntime version conflicts, CPU vs GPU variant mix-ups (only…

Inspect a target ONNX model and prepare metadata for Quark ONNX PTQ planning. Use when the user needs `.onnx` path validation, opset / IR version detection, input/output shape and dtype discovery, op-type histogram, quantizable-op counting, deployment-target…

Build a Quark ONNX PTQ quantization plan from `model_analysis.json` and user intent. Use when the user needs preset selection (XINT8 / A8W8 / A16W8 / BF16 / BFP16 / MX* / MXFP* …), calibration method choice (MinMax / Entropy / Percentile / Distribution /…

Validate Quark ONNX quantization output using four lightweight checks: auxiliary file copy alignment, expected non-quantized initializer MD5 byte-identity (inline `raw_data` + external-data byte ranges), model metadata equality after stripping…

Route Quark ONNX user goals to the correct atomic skill. Use when a user describes an ONNX quantization task in plain language — such as "install onnxruntime", "analyze my .onnx model", "choose a preset for my YOLO model", "plan ONNX PTQ", "quantize this…

Orchestrates modular PyTorch profiler trace analysis with TraceLens: generates perf reports, prepares category data, runs system-level and compute-kernel subagents in parallel, validates outputs, and writes a prioritized stakeholder report (analysis.md). Use…

Turns a natural-language description of routing intent into a valid Lemonade `collection.router` policy JSON. The skill generates and validates the JSON only - it does not register it or call the live server. Use when the user wants to route requests between…

Integrates local AI capabilities into applications using Embeddable Lemonade. Use when the user wants to add local AI, offline AI, private AI, on-device AI, a local LLM, local chat, embeddings, image generation, speech-to-text, or text-to-speech to an…

Makes this agent generate images, transcribe audio, and synthesize speech on the user's own machine through a local Lemonade Server instead of a paid cloud API. Use it above all to change that routing persistently, from now on — keep generating pictures…

Serves an LLM on a supported AMD EPYC server CPU using vLLM with zentorch, in Docker, Podman, or conda. Use for "vLLM on CPU", "zentorch serving", or an EPYC CPU endpoint, including on a host that also has AMD Instinct GPUs. Detects the EPYC generation,…

Autonomously optimizes end-to-end LLM inference throughput on AMD Instinct GPUs and reports a validated gain, using the Hyperloom multi-agent optimizer. Given a model, framework, workload (TP/EP, concurrency, ISL/OSL, precision), an objective and a time…

Benchmarks LLM inference and drives GPU kernel optimization with Magpie. Use when the user wants to benchmark vLLM, SGLang, or Atom; capture torch traces; post-process inference traces with TraceLens into prefill/decode and roofline reports; identify top…

Serves AI models on AMD Instinct GPU hardware using vLLM. Use this skill whenever the user wants to run, serve, deploy, start, host, or launch a language model on an AMD GPU, AMD Instinct, MI300X, MI325X, MI350X, or MI355X. Also use when the user mentions…

Measures vLLM serving performance on the zentorch/ZenDNN CPU backend on AMD EPYC. Brings up a multi-instance stack -- N cpuset-pinned `vllm serve` containers, one core slice each, behind an NGINX round-robin load balancer -- and drives it with GuideLLM…

Install vLLM plus the zentorch plugin against the ZenDNN backend. vLLM is obtained either from a pre-built CPU wheel (pip) or built from source — always ask the user which. If the user provides vLLM PR(s), force a source build and cherry-pick them. Always…

Build or rebuild the zentorch wheel from source (cmake auto-fetches ZenDNN), install it, and verify the version and build config. Use when the user asks to build, rebuild, or compile zentorch from source.

Prepare the Python environment (install the pinned PyTorch CPU build) and/or do a fresh end-to-end zentorch setup: validate PyTorch, install dependencies, build, install, and verify zentorch. Use when the user asks to create or prepare an environment, install…

Lint zentorch Python (flake8), C++ (git clang-format), or shell (shellcheck) code. Use when the user asks to lint, format-check, or run style checks on zentorch code.

Run zentorch tests (unit, op, model, miscellaneous, export, vLLM, LLM, or pre-trained model tests) with the required ZenDNN cache env vars. Use when the user asks to run or discover zentorch tests.

Build vLLM (CPU) end-to-end with the oneDNN + ZenDNN (zen64) backend via direct integration (NO zentorch). Use when asked to build vLLM CPU against oneDNN/ZenDNN or produce a ZenDNN-enabled vLLM wheel. Orchestrates the ZenDNN native lib (via the build-zendnn…

Build the standalone ZenDNN native library (zendnnl) from source with the alternate compute backends OFF (no oneDNN, libxsmm, parlooper, fbgemm), keeping AOCL DLP, which is the GEMM backend zendnnl links against. Use when asked to build ZenDNN alone, produce…

Build and verify the TensorFlow-ZenDNN plug-in (zentf) from source for the Python, C++, or Java interface on AMD EPYC CPUs. ALWAYS ask the user for the target TensorFlow version FIRST — it drives the whole build (which tensorflow is pip-installed, which…

Ports NVIDIA CUDA codebases to AMD HIP/ROCm end-to-end: runs hipify-perl, applies proactive wave-size and cooperative-groups fixes, maps libraries (cuBLAS→hipBLAS, cuFFT→hipFFT, cuDNN→MIOpen), configures Docker and ROCm for the target GPU arch, and migrates…