ML Ops
Machine learning operations
pytorch-distributed
by cuba6112
Distributed training strategies including DistributedDataParallel (DDP) and Fully Sharded Data Parallel (FSDP). Covers multi-node setup, checkpointing, and process management using torchrun. (ddp, fsdp, distributeddataparallel, torchrun, nccl, rank, process-group)
qwen-tts
by paki81
Local text-to-speech using Qwen3-TTS-12Hz-1.7B-CustomVoice. Use when generating audio from text, creating voice messages, or when TTS is requested. Supports 10 languages including Italian, 9 premium speaker voices, and instruction-based voice control (emotion, tone, style). Alternative to cloud-based TTS services like ElevenLabs. Runs entirely offline after initial model download.
Model Export & Serving
by sovr610
This skill should be used when the user asks to "export the model", "convert to ONNX", "trace with TorchScript", "quantize the model", "deploy the model", "create API server", "serve predictions", "containerize the model", "optimize for inference", "export to INT8", "create FastAPI endpoint", "set up model serving", "prune the model", "distill knowledge", "create Docker image", or needs guidance on model export formats, quantization strategies, serving infrastructure, or deployment pipelines for the brain_ai system.
data-science-feature-engineering
by legout
"Feature engineering for machine learning: encoding, scaling, transformations, datetime features, text features, and feature selection. Use when preparing data for modeling or improving model performance through better representations."
quality-documentation-manager
by nimeshgurung
Senior Quality Documentation Manager for comprehensive documentation control and regulatory document review. Provides document management system design, change control, configuration management, and regulatory documentation oversight. Use for document control system implementation, regulatory document review, change management, and documentation compliance verification.
everclaw
by profbernardoj
Open-source first AI inference — GLM-5 as default, Claude as fallback only. Own your inference forever via the Morpheus decentralized network. Stake MOR tokens, access GLM-5, GLM-4.7 Flash, Kimi K2.5, and 30+ models with persistent inference by recycling staked MOR. Open-source first model router routes all tiers to Morpheus by default — Claude only kicks in as an escape hatch when needed. Includes Morpheus API Gateway bootstrap for zero-config startup, OpenAI-compatible proxy with auto-session management, automatic retry with fresh sessions, OpenAI-compatible error classification to prevent cooldown cascades, multi-key auth rotation v2 with proactive DIEM balance monitoring and reactive 402 watchdog, Gateway Guardian v5 with direct curl inference probes (eliminates Signal spam), proactive Venice DIEM credit monitoring, circuit breaker for stuck sub-agents, nuclear self-healing restart, always-on proxy-router with launchd auto-restart, smart session archiver, three-shift cyclic execution engine (v2 with 15-minute execution loops), 24/7 always-on power configuration for macOS, bundled security skills, zero-dependency wallet management via macOS Keychain, x402 payment client for agent-to-agent USDC payments, ERC-8004 agent registry reader for discovering trustless agents on Base, and hardware-aware local Ollama fallback with auto model selection (Qwen3.5 family, 1.5B–72B based on available RAM/GPU).
mistral-model-ft-playbook
by Haruk1y
Build, fine-tune, evaluate, and ship Mistral-family models with Hugging Face and PEFT, with emphasis on strict JSON outputs and model-format compatibility. Use when choosing Mistral model variants, deciding prompting vs fine-tuning, preparing {prompt, completion} or chat datasets, running SFT jobs, validating generation failures, and deciding adapter-only vs merged full-model distribution.
prompt-caching
by anilsoylu
Guide for optimizing prompt caching in agentic AI applications. Use when building Claude agents, optimizing API costs, designing system prompts, managing tool definitions, implementing compaction, or troubleshooting cache miss rates. Triggers: "prompt caching", "cache hit rate", "reduce API costs", "system prompt ordering", "tool definitions caching", "compaction", "context window management", "agent cost optimization", "prefix caching", "cache-safe forking", "agentic loop cost", "token cost reduction", "cache miss debugging".
qa
by pmco23
Use after /build to run the full post-build QA pipeline. Supports --parallel (all audits simultaneously) or --sequential (denoise → qf → qb → qd → security-review in order). Requires .pipeline/build.complete.
create-agent-skills
by myuon
新しい Claude Code スキルを作成するためのガイド。スキルの作り方、SKILL.md の書き方、配置場所を案内する。
model-router
by cgk-platform
Cost-optimized bi-directional model routing. Downgrade routine tasks to Flash, upgrade complex tasks to Opus.
eve-local-dev-loop
by Incept5
Local Docker Compose development loop for Eve-compatible apps, with handoff to staging deploys.
AI Security Engineer
by anorbert-cmyk
AI Security Engineer focused on preventing prompt injection, data exfiltration, and model abuse in LLM-enabled products.
SystematicNotes —— 自动化课程笔记生成系统
by CoEvolution-Labs
自动化课程资料整理与笔记生成系统 — 把 PPT + 录音 → 图文并茂的精品 Markdown 笔记
Brand Name Comparison
by MaxGhenis
Use this skill when the user wants to compare multiple brand name fates side-by-side to divine which is destined for success.
browser-onnx
by thongnt0208
"Implements high-performance local machine learning inference in the browser using ONNX Runtime Web. Use this skill when the user needs privacy-first, low-latency, or offline AI capabilities (e.g., image classification, object detection, or NLP) without server-side processing."
Continual Learning Guard (EWC / SI / Progressive Networks / Replay)
by sovr610
This skill should be used when the user asks to "prevent catastrophic forgetting", "elastic weight consolidation", "EWC regularization", "progressive networks", "continual learning strategy", "knowledge distillation for retention", "replay buffer memory", "task boundary detection", "fisher information matrix", "synaptic intelligence", "PackNet pruning", "memory-aware synapses", "add continual learning guard", "implement EWC penalty", "add experience replay", "implement progressive columns", "add Fisher diagonal computation", "implement reservoir sampling", "add knowledge distillation loss", "implement task-free continual learning", "add online EWC", "implement generative replay", "add PackNet iterative pruning", "implement synaptic intelligence path integral", or mentions catastrophic forgetting, continual learning, lifelong learning, sequential task training, knowledge retention, task interference, Fisher information regularization, or Phase 8 continual-learning pipeline in the cognitive architecture.
m09-domain
by lywa1998
"CRITICAL: Use for domain modeling. Triggers: domain model, DDD, domain-driven design, entity, value object, aggregate, repository pattern, business rules, validation, invariant, 领域模型, 领域驱动设计, 业务规则"
wandb-weave-ft-retrospective
by Haruk1y
Analyze W&B, Weave, and local fine-tuning evaluation artifacts, then produce a concrete next-run improvement plan with data, prompt, and training actions. Use after each SFT or eval cycle.
ai-image-generation
by Jackiexiao
"Generate AI images with FLUX, Gemini, Grok, Seedream, Reve and 50+ models via inference.sh CLI. Models: FLUX Dev LoRA, FLUX.2 Klein LoRA, Gemini 3 Pro Image, Grok Imagine, Seedream 4.5, Reve, ImagineArt. Capabilities: text-to-image, image-to-image, inpainting, LoRA, image editing, upscaling, text rendering. Use for: AI art, product mockups, concept art, social media graphics, marketing visuals, illustrations. Triggers: flux, image generation, ai image, text to image, stable diffusion, generate image, ai art, midjourney alternative, dall-e alternative, text2img, t2i, image generator, ai picture, create image with ai, generative ai, ai illustration, grok image, gemini image"
pipeline-pattern-nodejs
by progmichaelkibenko
Implements the Pipeline design pattern in Node.js for data transformation. Use when the user mentions pipeline pattern, or when you need a fixed sequence of stages that each transform data and pass to the next—ETL, parsing, data processing, or any linear transformation flow that runs to completion.
Active Inference Agent (Generative Model + EFE + Empowerment)
by sovr610
This skill should be used when the user asks to "implement active inference", "add EFE computation", "implement expected free energy", "add empowerment estimation", "implement generative model", "add latent state encoder", "implement transition model", "add preference model", "implement planning rollouts", "add CEM planner", "implement amortized policy", "add pymdp backend", "implement offline RL", "add Minari integration", "implement pragmatic value", "add epistemic value", "implement instrumental value", "add action selection", "implement belief updating", "add world model training", "implement POMDP planning", "add rollout engine", "implement latent imagination", "add horizon normalization", "implement cross-entropy method planning", "add preference learning", "implement variational empowerment", or mentions active inference, expected free energy decomposition, POMDP planning, empowerment estimation, latent imagination, or decision-as-inference in the cognitive pipeline.
mech-interp
by Pranav-Karra-3301
Mechanistic interpretability skill for understanding neural network internals. Use when: (1) Working with TransformerLens, nnsight, or SAELens (2) Analyzing attention patterns, circuits, or features (3) Training or analyzing sparse autoencoders (SAEs) (4) Activation patching, logit lens, or direct logit attribution (5) Understanding model internals ("what is this head doing?") (6) Setting up a mech interp research project Triggers: "mech interp", "mechanistic interpretability", "TransformerLens", "SAE", "sparse autoencoder", "activation patching", "logit lens", "circuits", "features", "attention heads", "residual stream", "nnsight", "nnterp", "superposition", "polysemantic", "monosemantic", "ablation", "probing", "induction head"
Compiler & Kernel Fusion (torch.compile) Integration
by sovr610
This skill should be used when the user asks to "enable torch.compile", "add compilation to training", "kernel fusion", "TorchDynamo integration", "TorchInductor optimization", "reduce-overhead mode", "max-autotune mode", "fix graph breaks", "compile health check", "shape stabilization", "dynamic shapes for compile", "bucketing for torch.compile", "compile allowlist", "compile blocklist", "compile smoketest", "CUDA graphs for training", "maybe_compile wrapper", "debug recompiles", "TORCH_LOGS compile", "compile + DDP", "compile + FSDP", "torch.compiler.disable", or needs guidance on torch.compile integration, shape management, compilation debugging, or safe fallback patterns.