Skills by huggingface

Dangerous skill: documents the `hf` CLI for Hugging Face Hub operations (download, upload, jobs, cache). It instructs executing shell commands (e.g., `hf jobs run <image> <command>`), network calls to `https://huggingface.co/...`, and sending secrets via `--env-file .env` and `--secrets HF_TOKEN`.

This skill manages Hugging Face datasets (repo init, template validation, streaming rows, DuckDB SQL queries, push/export). It requires `HF_TOKEN`, instructs running CLI commands like `uv run scripts/sql_manager.py ...`, and performs network access to `hf://`/`https://huggingface.co`.

High-risk skill: automates TRL training and GGUF conversion on Hugging Face Jobs with `hf_jobs`, Trackio integration, and model push. It requests `HF_TOKEN`, makes network calls to `https://github.com/ggerganov/llama.cpp.git` and `https://datasets-server.huggingface.co`, and runs system commands like `apt-get`/`cmake` to build and execute binaries.

This skill manages creation, indexing, and linking of research papers on Hugging Face Hub and provides CLI scripts and templates. It performs network requests to `https://huggingface.co/papers/{arxiv-id}` and `http://export.arxiv.org/api/query`, reads the `HF_TOKEN` environment variable, and instructs running `python scripts/paper_manager.py` commands.

This skill extracts/imports evaluation results and updates Hugging Face model cards, and can run local or HF Jobs evaluations via `uv`/`hf jobs`. It reads `HF_TOKEN` and `AA_API_KEY`, makes network calls to `https://artificialanalysis.ai` and `https://huggingface.co`, and runs subprocesses (e.g., `subprocess.run`, `lighteval`, `inspect`).

Track and visualize ML training experiments with Trackio. Use when logging metrics during training (Python API) or retrieving/analyzing logged metrics (CLI). Supports real-time dashboard visualization, HF Space syncing, and JSON output for automation.

This skill builds command-line scripts and utilities to query and process Hugging Face model and dataset data. It runs shell/Python tools and makes network calls to `https://huggingface.co`, and reads the `HF_TOKEN` environment variable for authentication.