Skills.
Give your AI a better way to work.
A skill is a set of written instructions that teaches an AI how to do one job the way it should be done: review a pull request, plan a migration, write the release notes.
Install one here and it travels with your account into Claude, Claude Code, Cursor and every other client you sign in with.
Filter by category
55,543 results · page 102 of 1,852
- View details
modal-serverless-gpuSkillCloud & infra
Lets your agent run ML workloads on on-demand cloud GPUs without managing servers.
Ready to connect★ 13k
- View details
model-mergingSkillAI & models
Lets your agent combine several trained AI models into one new model without retraining.
Ready to connect★ 13k
- View details
model-pruningSkillAI & models
Lets your agent shrink large language models for faster, lighter inference without retraining.
Ready to connect★ 13k
- View details
moe-trainingSkillAI & models
Lets your agent train sparse Mixture of Experts AI models with less compute using DeepSpeed or HuggingFace.
Ready to connect★ 13k
- View details
nanogptSkillAI & models
Lets your agent read and explain a small, clean GPT implementation for learning how transformers work.
Ready to connect★ 13k
- View details
nemo-curatorSkillDatabases & data
Lets your agent clean and filter large text, image, video, or audio datasets for AI model training.
Ready to connect★ 13k
- View details
nemo-guardrailsSkillAI & models
Lets your agent check AI inputs and outputs for jailbreaks, toxic content, personal data, and false claims.
Ready to connect★ 13k
- View details
nnsight-remote-interpretabilitySkillAI & models
Lets your agent run interpretability experiments on huge AI models remotely, without needing a local GPU.
Ready to connect★ 13k
- View details
openrlhf-trainingSkillAI & models
Lets your agent fine-tune large language models using RLHF methods like PPO, GRPO, RLOO, and DPO.
Ready to connect★ 13k
- View details
optimizing-attention-flashSkillDocs & knowledge
Helps your agent speed up transformer attention and cut GPU memory use when running models with long sequences.
Ready to connect★ 13k
- View details
outlinesSkillAI & models
Lets your agent generate AI model outputs that always follow a strict JSON, XML, or code structure.
Ready to connect★ 13k
- View details
peft-fine-tuningSkillDocs & knowledge
Lets your agent customize large language models by training a tiny fraction of their parameters on limited GPU memory.
Ready to connect★ 13k
- View details
phoenix-observabilitySkillDatabases & data
Lets your agent trace, evaluate, and monitor LLM applications to debug issues and track performance.
Ready to connect★ 13k
- View details
pineconeSkillSearch
Lets your agent store and search documents by meaning using a managed vector database.
Ready to connect★ 13k
- View details
presenting-conference-talksSkillDocs & knowledge
Lets your agent turn a research paper into conference slides with speaker notes and a talk script.
Ready to connect★ 13k
- View details
prompt-guardSkillCloud & infra
Lets your agent screen incoming prompts and third-party text for injection attacks or jailbreak attempts.
Ready to connect★ 13k
- View details
pytorch-fsdp2SkillDocs & knowledge
Lets your agent add PyTorch FSDP2 sharding to training scripts so large models fit across multiple GPUs.
Ready to connect★ 13k
- View details
pyvene-interventionsSkillMonitoring & ops
Guides your agent in running causal experiments on PyTorch models, like tracing which parts cause a behavior.
Ready to connect★ 13k
- View details
qdrant-vector-searchSkillSearch
Lets your agent search stored documents by meaning to find relevant information for answers.
Ready to connect★ 13k
- View details
quantizing-models-bitsandbytesSkillDocs & knowledge
Guides your agent through compressing large language models to 8-bit or 4-bit so they use much less GPU memory.
Ready to connect★ 13k
- View details
ray-dataSkillDatabases & data
Lets your agent process large datasets for machine learning, like converting files or running batch predictions, across many machines.
Ready to connect★ 13k
- View details
ray-trainSkillAI & models
Lets your agent launch and manage machine learning training jobs across many machines, with automatic tuning and recovery.
Ready to connect★ 13k
- View details
rwkv-architectureSkillProductivity
Lets your agent explain the RWKV model architecture, a hybrid design with linear-time inference and unbounded context.
Ready to connect★ 13k
- View details
segment-anything-modelSkillMedia
Lets your agent outline and separate objects in images using points, boxes, or masks as prompts.
Ready to connect★ 13k
- View details
sentence-transformersSkillSearch
Lets your agent turn text or images into embeddings for semantic search, similarity, and clustering tasks.
Ready to connect★ 13k
- View details
sentencepieceSkillDocs & knowledge
Lets your agent break text into subword pieces using BPE or Unigram, including multilingual and CJK text.
Ready to connect★ 13k
- View details
serving-llms-vllmSkillDocs & knowledge
Lets your agent serve LLMs as an API server that handles many requests on limited GPU memory.
Ready to connect★ 13k
- View details
sglangSkillAI & models
Lets your agent run language models faster and get structured outputs like JSON or regex-constrained text.
Ready to connect★ 13k
- View details
simpo-trainingSkillAI & models
Lets your agent fine-tune language models to match human preferences using a simpler, faster method than DPO.
Ready to connect★ 13k
- View details
skypilot-multi-cloud-orchestrationSkillDev tools
Lets your agent run machine learning training and batch jobs across multiple cloud providers while cutting GPU costs.
Ready to connect★ 13k
What is a skill?
A skill is plain text, usually a SKILL.md file and the scripts it refers to, written for an AI rather than for a person. It carries the steps, the house rules and the examples a good answer needs, so you stop pasting the same briefing into every new chat.
55,111 of the 55,543 skills listed here can be served through ahel today, and they come from public repositories. Each one has its own page with the instructions themselves on it, so you can read what a skill will tell your AI to do before you install it.
Install one and every AI you use gets it
Installing a skill adds it to your gateway and turns it on in the same step. Claude Code surfaces it as a slash command; any client can read the full instructions with the skill_read tool.
Nothing is copied into a project folder. The instructions are served from your account, so the same skill is there in every AI you connect, and turning it off removes it from all of them at once.