Skills

Introduction

Skills are plugins for AI coding agents that provide domain-specific knowledge and capabilities. They are distributed through the Datathings marketplace and can be installed in Claude Code using the /plugin command.

Setup

Add the marketplace:

/plugin marketplace add datathings/marketplace

Install a skill:

/plugin install greycat@datathings

Available skills

Skill	Description	Source
GreyCat
greycat	Full-stack GreyCat development — GCL language, graph persistence, frontend integration, LSP	SKILL.md
greycat-c	GreyCat C API and standard library for native development	SKILL.md
Agentic AI
llamacpp	llama.cpp C API reference (163 functions) for local LLM inference	SKILL.md
ggml	ggml C tensor library (560+ functions) for graph computation, GGUF I/O, multi-backend inference	SKILL.md
vllm	vLLM v0.16.0 — high-throughput Python LLM inference with batch, OpenAI-compatible server, LoRA, structured outputs	SKILL.md
ollama	Ollama v0.16.3 — run local LLMs via REST API (chat, generate, embed, model management)	SKILL.md
High Performance Math & GPU
blas_lapack	CBLAS & LAPACKE C API reference (1284 functions) for numerical linear algebra	SKILL.md
cuda	NVIDIA CUDA C/C++ — Runtime API, cuBLAS, cuFFT, cuSPARSE, cuRAND, cuSolver, Thrust	SKILL.md
opencl	OpenCL SDK (Khronos) — cross-platform GPU/CPU parallel computing, C API, C++ wrapper	SKILL.md
rocm	AMD ROCm 7.2.0 — HIP kernel development, rocBLAS/rocFFT/rocRAND/rocSOLVER libraries	SKILL.md
Power Grid
pandapower	pandapower v3.4.0 — AC/DC power flow, OPF, short circuit (IEC 60909), state estimation	SKILL.md
powergridmodel	power-grid-model v1.13.10 — steady-state distribution power system analysis	SKILL.md

Skill details

greycat

Source

Activates on .gcl files and GreyCat topics. Provides:

• GCL syntax, types, decorators (@expose, @permission, @volatile)
• Indexed collections (nodeIndex, nodeList, nodeTime, nodeGeo)
• Concurrency patterns (Jobs, await)
• Standard library (core, io, runtime, util)
• Pro libraries (ai, algebra, finance, kafka, opcua, powerflow, s3, sql, useragent)
• Frontend integration (@greycat/web)

/plugin install greycat@datathings

greycat-c

Source

Reference for native C development with GreyCat:

• C API functions, tensor operations
• Native function implementation patterns

/plugin install greycat-c@datathings

llamacpp

Source

Complete llama.cpp C API (163 functions):

• Model loading, inference, tokenization
• Sampling strategies (XTC, DRY, infill)
• GGUF model support

/plugin install llamacpp@datathings

ggml

Source

C tensor computation library powering llama.cpp (v0.9.7, 560+ functions):

• Lazy computation graph with CPU/GPU/Metal/Vulkan backends and automatic multi-backend scheduling
• 40+ quantization formats (Q4_0 → Q5_K), GGUF v3 I/O, Flash Attention, RoPE, AdamW/SGD training

/plugin install ggml@datathings

vllm

Source

vLLM (v0.16.0) — high-throughput Python inference engine for large language models:

• Offline batch inference (LLM class) and OpenAI-compatible server (vllm serve) with streaming
• LoRA adapters, multimodal inputs, structured outputs (JSON/regex/grammar), paged attention

/plugin install vllm@datathings

ollama

Source

Ollama (v0.16.3) — local LLM runtime with a REST API:

• Text generation, chat, and embeddings with streaming support
• Model management (pull/push/delete) and custom model creation via Modelfile

/plugin install ollama@datathings

blas_lapack

Source

Complete CBLAS & LAPACKE C API (1284 functions, LAPACK v3.12.1):

• BLAS Level 1/2/3 vector and matrix operations; linear solvers (LU, Cholesky, LDL)
• Eigenvalue/SVD/least squares decompositions; QR/LQ factorizations

/plugin install blas_lapack@datathings

cuda

Source

NVIDIA CUDA parallel computing platform (cuda-samples v13.1):

• Runtime API (device, memory, streams, kernel launch); cuBLAS, cuFFT, cuSPARSE, cuRAND, cuSolver
• Thrust (STL-like GPU algorithms) and Cooperative Groups (thread synchronization)

/plugin install cuda@datathings

opencl

Source

Khronos Group OpenCL SDK (v2025.07.23) for cross-platform GPU/CPU parallel computing:

• C API (~60 functions) and C++ wrapper (opencl.hpp with RAII types)
• Full NDRange execution (1D/2D/3D), events, profiling, out-of-order queues

/plugin install opencl@datathings

rocm

Source

AMD ROCm GPU computing stack (v7.2.0) for HIP-based development:

• HIP C++ kernels with compute libraries (rocBLAS, rocFFT, rocRAND, rocSOLVER, rocSPARSE, rocWMMA)
• Profiling (rocProfiler, rocm-smi) and CUDA portability via hipify-perl

/plugin install rocm@datathings

pandapower

Source

pandapower (v3.4.0) — Python library for electric power network analysis:

• AC/DC power flow, optimal power flow, short-circuit (IEC 60909), state estimation
• 15+ benchmark networks (IEEE, CIGRE, Kerber) and visualization with matplotlib/plotly

/plugin install pandapower@datathings

powergridmodel

Source

power-grid-model (v1.13.10) — high-performance Python/C++ library for distribution power system analysis:

• Symmetric and asymmetric three-phase power flow, state estimation, IEC 60909 short-circuit
• Batch/N-1 contingency analysis with multi-threaded parallel execution; 22 component types

/plugin install powergridmodel@datathings

Configuration

Skills can also be configured manually in .claude/settings.json (project-level or ~/.claude/settings.json for global):

{
  "extraKnownMarketplaces": {
    "datathings": {
      "source": { "source": "github", "repo": "datathings/marketplace" }
    }
  },
  "enabledPlugins": {
    "greycat@datathings": true
  }
}

Management

/plugin list                              # List installed plugins
/plugin update greycat@datathings         # Update a plugin
/plugin uninstall greycat@datathings      # Remove a plugin
/plugin marketplace list                  # List marketplaces
/plugin marketplace remove datathings     # Remove marketplace