Huggingface Best
插件 活跃Find the best AI model for any task by querying Hugging Face leaderboards and benchmarks. Recommends top models based on task type, hardware constraints, and benchmark scores.
To help users find the most suitable AI model for their specific tasks and hardware limitations by leveraging Hugging Face's benchmark data.
功能
- Query Hugging Face leaderboards and benchmarks
- Recommend top models based on task type
- Filter models by hardware constraints (e.g., device memory)
- Enrich model data with parameter count and license information
- Present model comparisons in a tabular format
使用场景
- When seeking the best AI model for a specific task like coding or text generation.
- When comparing different AI models based on benchmark scores and size.
- When unsure which AI model can run on available hardware.
- When needing to quickly identify state-of-the-art models for a given use case.
非目标
- Training or fine-tuning AI models.
- Running AI models directly.
- Providing an exhaustive list of all available models.
- Evaluating models not present on official Hugging Face leaderboards.
Security
- warning:Secret ManagementThe skill script references `~/.cache/huggingface/token` and `$(cat ~/.cache/huggingface/token)` for API access, suggesting it might expose or handle user tokens without clear indication of secure handling.
- info:Data ExfiltrationThe skill reads a Hugging Face token from a local file for API authentication, which is a common practice but should ideally be handled more securely.
- warning:Keychain-stored secretsThe skill references a Hugging Face token from `~/.cache/huggingface/token`, which is likely stored in plain text and not routed through secure keychain storage.
Code Execution
- warning:ValidationThe skill parses user input for task and device, but it does not appear to use a schema validation library for input sanitization or parameter validation.
- warning:Error HandlingThe skill includes basic error handling for API calls (e.g., 'leaderboard not found'), but it lacks structured error reporting with retryable flags or hints for the agent.
Errors
- warning:Actionable error messagesError messages for API call failures are present but lack specific remediation steps or links to documentation, making them less actionable for the agent.
安装
请先添加 Marketplace
/plugin marketplace add huggingface/skills/plugin install huggingface-best@huggingface-skills质量评分
类似扩展
Context7 Plugin
100Upstash Context7 MCP 服务器，用于查找最新文档。直接从源代码存储库将特定版本的文档和代码示例拉取到您的 LLM 上下文中。
Hugging Face Papers
100Look up and read Hugging Face paper pages in markdown, and use the papers API for structured metadata like authors, linked models, datasets, Spaces, and media URLs when needed.
Huggingface Trackio
99Track and visualize ML training experiments with Trackio. Log metrics via Python API and retrieve them via CLI. Supports real-time dashboards synced to HF Spaces.
Hf Cli
99Execute Hugging Face Hub operations using the hf CLI. Download models/datasets, upload files, manage repos, and run cloud compute jobs.
Huggingface Local Models
99Use to select models to run locally with llama.cpp and GGUF on CPU, Mac Metal, CUDA, or ROCm. Covers finding GGUFs, quant selection, running servers, exact GGUF file lookup, conversion, and OpenAI-compatible local serving.
Huggingface Llm Trainer
99Train or fine-tune language models using TRL on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes hardware selection, cost estimation, Trackio monitoring, and Hub persistence.