[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"extension-plugin-huggingface-huggingface-llm-trainer-zh-CN":3,"guides-for-huggingface-huggingface-llm-trainer":744,"similar-k175sdyyg60kjnnbkm4029s70186ngbn-zh-CN":745},{"_creationTime":4,"_id":5,"children":6,"community":7,"display":9,"evaluation":14,"identity":252,"isFallback":249,"parentExtension":256,"providers":291,"relations":295,"repo":296,"tags":742,"workflow":743},1778690773482.4827,"k175sdyyg60kjnnbkm4029s70186ngbn",[],{"reviewCount":8},0,{"description":10,"installMethods":11,"name":12,"sourceUrl":13},"Train 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.",{"claudeCode":12},"huggingface-llm-trainer","https://github.com/huggingface/skills",{"_creationTime":15,"_id":16,"extensionId":5,"locale":17,"result":18,"trustSignals":233,"workflow":250},1778690836703.0266,"kn7462pfdj5xnhc2dyasx1dq7186nvtr","en",{"checks":19,"evaluatedAt":200,"extensionSummary":201,"features":202,"nonGoals":208,"promptVersionExtension":212,"promptVersionScoring":213,"purpose":214,"rationale":215,"score":216,"summary":217,"tags":218,"targetMarket":226,"tier":227,"useCases":228},[20,25,28,31,35,38,42,46,50,53,57,61,64,68,71,74,77,80,83,86,90,94,98,102,106,109,112,115,119,122,125,128,131,134,137,141,145,149,152,156,159,162,165,168,171,174,177,180,183,186,190,193,196],{"category":21,"check":22,"severity":23,"summary":24},"Practical Utility","Problem relevance","pass","The description clearly states the problem of training or fine-tuning language models using TRL on Hugging Face Jobs infrastructure, covering specific methods and local deployment.",{"category":21,"check":26,"severity":23,"summary":27},"Unique selling proposition","The skill offers significant value by abstracting complex Hugging Face Jobs infrastructure, TRL training methods, GGUF conversion, and related tooling into a single, coherent interface for AI agents.",{"category":21,"check":29,"severity":23,"summary":30},"Production readiness","The skill covers the complete lifecycle of LLM training, from hardware selection and cost estimation to training execution, monitoring, and GGUF conversion for local deployment.",{"category":32,"check":33,"severity":23,"summary":34},"Scope","Single responsibility principle","The skill focuses on a coherent workflow of language model training and fine-tuning using TRL and Hugging Face infrastructure, including related utilities like GGUF conversion and cost estimation.",{"category":32,"check":36,"severity":23,"summary":37},"Description quality","The provided description accurately reflects the skill's capabilities, including the various training methods, infrastructure usage, and supporting features.",{"category":39,"check":40,"severity":23,"summary":41},"Invocation","Scoped tools","The skill utilizes the `hf_jobs()` MCP tool with specific parameters and scripts, indicating a scoped approach rather than a general-purpose command executor.",{"category":43,"check":44,"severity":23,"summary":45},"Documentation","Configuration & parameter reference","The SKILL.md thoroughly documents prerequisites, configuration options for training jobs (timeout, secrets, hardware), and provides detailed explanations for various approaches and parameters.",{"category":32,"check":47,"severity":48,"summary":49},"Tool naming","not_applicable","This skill primarily utilizes an MCP tool (`hf_jobs`) and documented approaches rather than exposing numerous distinct tools with names that would require evaluation.",{"category":32,"check":51,"severity":23,"summary":52},"Minimal I/O surface","The skill's core interaction is via the `hf_jobs` MCP tool, which is designed for structured input and output. The SKILL.md also guides on providing necessary parameters, implying a controlled I/O.",{"category":54,"check":55,"severity":23,"summary":56},"License","License usability","The extension is licensed under the Apache-2.0 license, as indicated by the bundled LICENSE file, which is a permissive open-source license.",{"category":58,"check":59,"severity":23,"summary":60},"Maintenance","Commit recency","The repository has recent commits as of May 2026, indicating active maintenance.",{"category":58,"check":62,"severity":23,"summary":63},"Dependency Management","The skill uses PEP 723 for inline dependencies, allowing for explicit versioning within scripts, and the repository has CI, suggesting dependency checks.",{"category":65,"check":66,"severity":23,"summary":67},"Security","Secret Management","Secrets like HF_TOKEN are handled via the `secrets` parameter in `hf_jobs`, which is designed for secure passing of sensitive information, and the documentation explicitly warns about the ephemeral nature of the environment.",{"category":65,"check":69,"severity":23,"summary":70},"Injection","The skill executes Python scripts or fetches them from trusted URLs, and the documentation emphasizes using scripts as data, not instructions, mitigating injection risks.",{"category":65,"check":72,"severity":23,"summary":73},"Transitive Supply-Chain Grenades","Scripts are either inline Python or fetched from Hugging Face Hub/GitHub URLs, and the use of PEP 723 dependencies within scripts mitigates risks from runtime external fetches.",{"category":65,"check":75,"severity":23,"summary":76},"Sandbox Isolation","Jobs run in isolated Docker containers managed by Hugging Face Jobs, ensuring operations are confined to the job environment and do not affect the user's local system.",{"category":65,"check":78,"severity":23,"summary":79},"Sandbox escape primitives","The skill relies on the Hugging Face Jobs infrastructure, which is designed with robust sandbox isolation, preventing detached processes or escape mechanisms.",{"category":65,"check":81,"severity":23,"summary":82},"Data Exfiltration","The skill's primary function is training, and it requires explicit secrets like HF_TOKEN only for pushing results to the Hub, with clear documentation on required permissions and no indication of undocumented outbound calls.",{"category":65,"check":84,"severity":23,"summary":85},"Hidden Text Tricks","The SKILL.md content is clean, standard markdown with no apparent hidden steering characters or formatting tricks.",{"category":87,"check":88,"severity":23,"summary":89},"Hooks","Opaque code execution","Scripts are either inline Python or fetched from known sources and executed via `hf_jobs`, which are designed for straightforward script execution without obfuscation.",{"category":91,"check":92,"severity":23,"summary":93},"Portability","Structural Assumption","The skill operates within the Hugging Face Jobs environment, abstracting away local project structure dependencies. Scripts are either inline or fetched from remote URLs.",{"category":95,"check":96,"severity":23,"summary":97},"Trust","Issues Attention","The repository shows 4 open and 6 closed issues in the last 90 days, with a closure rate of 60%, indicating active maintainer engagement.",{"category":99,"check":100,"severity":23,"summary":101},"Versioning","Release Management","The repository has recent commit activity, and the presence of a LICENSE file and clear documentation on usage implies a versioned and managed extension.",{"category":103,"check":104,"severity":23,"summary":105},"Code Execution","Validation","The SKILL.md emphasizes dataset validation using `dataset_inspector.py` before training, and the use of PEP 723 dependencies provides a form of validation for script requirements.",{"category":65,"check":107,"severity":23,"summary":108},"Unguarded Destructive Operations","The primary operations involve submitting jobs to Hugging Face, which are managed through a controlled environment. Destructive actions like data deletion are not part of the core training workflow.",{"category":103,"check":110,"severity":23,"summary":111},"Error Handling","The SKILL.md details how Hugging Face Jobs handle errors, including timeouts and dataset format issues, and provides guidance on interpreting and resolving them. The use of Python scripts within `hf_jobs` implies standard Python error handling.",{"category":103,"check":113,"severity":23,"summary":114},"Logging","The SKILL.md explicitly mentions Trackio for monitoring and provides commands for `hf_jobs` to view logs, enabling users to review actions and outcomes.",{"category":116,"check":117,"severity":48,"summary":118},"Compliance","GDPR","The skill focuses on model training and does not appear to handle personal data directly. Any data submitted to the LLM for training is user-provided and within the scope of the training task.",{"category":116,"check":120,"severity":23,"summary":121},"Target market","The extension is global in scope, focusing on AI model training infrastructure accessible worldwide, with no specific regional limitations mentioned.",{"category":91,"check":123,"severity":23,"summary":124},"Runtime stability","The skill operates within the Hugging Face Jobs environment, which is designed to be platform-agnostic. Scripts use standard Python and PEP 723 dependencies.",{"category":43,"check":126,"severity":23,"summary":127},"README","The README clearly states the purpose of Hugging Face Skills and provides installation and usage instructions for various agents.",{"category":32,"check":129,"severity":48,"summary":130},"Tool surface size","This is a plugin-level evaluation; the skill itself primarily interfaces via the `hf_jobs` MCP tool, not a large set of individual commands.",{"category":39,"check":132,"severity":48,"summary":133},"Overlapping near-synonym tools","As a plugin skill, it primarily uses the `hf_jobs` MCP tool, avoiding direct exposure of multiple overlapping tools.",{"category":43,"check":135,"severity":23,"summary":136},"Phantom features","All described features, such as SFT, DPO, GRPO training, GGUF conversion, and cost estimation, are directly supported by the `hf_jobs` tool and documented examples.",{"category":138,"check":139,"severity":23,"summary":140},"Install","Installation instruction","The README provides clear installation instructions for Claude Code, Codex, Gemini CLI, and Cursor, including copy-pasteable commands and setup guidance.",{"category":142,"check":143,"severity":23,"summary":144},"Errors","Actionable error messages","The SKILL.md provides detailed guidance on common error scenarios like OOM, dataset format mismatches, timeouts, and Hub push failures, including remediation steps.",{"category":146,"check":147,"severity":23,"summary":148},"Execution","Pinned dependencies","The use of PEP 723 inline dependencies within scripts ensures that dependencies are declared and can be managed explicitly for each script.",{"category":32,"check":150,"severity":23,"summary":151},"Dry-run preview","While not a direct `--dry-run` flag for the entire skill, the `scripts/estimate_cost.py` tool provides a preview of training time and cost, and the `dataset_inspector.py` validates dataset format beforehand.",{"category":153,"check":154,"severity":23,"summary":155},"Protocol","Idempotent retry & timeouts","Hugging Face Jobs handle task execution and retries. The skill documentation emphasizes setting appropriate timeouts for training jobs.",{"category":116,"check":157,"severity":23,"summary":158},"Telemetry opt-in","The skill promotes Trackio for monitoring, which is explicitly integrated via `report_to='trackio'` and requires user configuration, implying an opt-in mechanism.",{"category":39,"check":160,"severity":48,"summary":161},"Name collisions","This is a single plugin skill, not a collection of multiple distinct extensions within a bundle that could have name collisions.",{"category":39,"check":163,"severity":48,"summary":164},"Hooks-off mechanism","This skill does not appear to utilize hooks in a way that would require a separate hooks-off mechanism.",{"category":39,"check":166,"severity":48,"summary":167},"Hook matcher tightness","The skill does not appear to rely on custom hooks with specific matchers; it uses the `hf_jobs` MCP tool.",{"category":65,"check":169,"severity":48,"summary":170},"Hook security","The skill does not appear to use hooks for potentially destructive or network-sensitive operations.",{"category":87,"check":172,"severity":48,"summary":173},"Silent prompt rewriting","The skill does not appear to have a `UserPromptSubmit` hook that would rewrite prompts.",{"category":65,"check":175,"severity":48,"summary":176},"Permission Hook","There is no indication of `PermissionRequest` hooks being used in this skill.",{"category":116,"check":178,"severity":48,"summary":179},"Hook privacy","The skill does not appear to implement hooks that involve logging or telemetry sent over the network.",{"category":103,"check":181,"severity":48,"summary":182},"Hook dependency","The skill does not appear to rely on custom hooks written in separate scripts.",{"category":43,"check":184,"severity":23,"summary":185},"Feature Transparency","The SKILL.md clearly describes the training methods, infrastructure, and supporting utilities, providing transparency into the skill's functionality.",{"category":187,"check":188,"severity":23,"summary":189},"Convention","Layout convention adherence","As a skill within a larger repository, it adheres to the expected structure for skills, with a dedicated SKILL.md and supporting scripts.",{"category":187,"check":191,"severity":48,"summary":192},"Plugin state","This skill does not manage persistent plugin state; its operations are primarily job-based and ephemeral.",{"category":65,"check":194,"severity":23,"summary":195},"Keychain-stored secrets","Secrets are handled via the `secrets` parameter in `hf_jobs`, which is the recommended mechanism for securely passing sensitive information to jobs.",{"category":197,"check":198,"severity":48,"summary":199},"Installation","Clean uninstall","This skill does not install background daemons or persistent processes, so a clean uninstall is implicit when the plugin is removed.",1778690836577,"This plugin provides a skill for training and fine-tuning language models using TRL and Hugging Face Jobs. It supports various training methods (SFT, DPO, GRPO), GGUF conversion, hardware selection, cost estimation, and monitoring via Trackio.",[203,204,205,206,207],"Train/fine-tune LLMs with TRL (SFT, DPO, GRPO)","Utilize Hugging Face Jobs for cloud GPU training","Convert models to GGUF for local deployment","Estimate training cost and hardware requirements","Integrate Trackio for real-time monitoring",[209,210,211],"Performing training directly on the user's local machine","Managing Hugging Face Hub repositories beyond saving trained models","Providing a UI for model training; all interactions are agent-driven","3.0.0","4.4.0","Enables AI agents to easily train and fine-tune large language models on powerful cloud infrastructure without requiring local setup or complex configuration.","Excellent documentation and adherence to best practices for Hugging Face Jobs and TRL training result in a high score. No critical or warning findings.",99,"Comprehensive and well-documented skill for advanced LLM training on Hugging Face infrastructure.",[219,220,221,222,223,224,225],"llm","training","fine-tuning","huggingface","trl","gguf","mlops","global","verified",[229,230,231,232],"Fine-tuning LLMs for specific tasks on cloud GPUs","Experimenting with different TRL training methods","Converting trained models to GGUF for local inference","Estimating the cost and time for LLM training jobs",{"codeQuality":234,"collectedAt":236,"documentation":237,"maintenance":240,"security":246,"testCoverage":248},{"hasLockfile":235},false,1778690814383,{"descriptionLength":238,"readmeSize":239},271,9821,{"closedIssues90d":241,"forks":242,"hasChangelog":235,"openIssues90d":243,"pushedAt":244,"stars":245},6,663,4,1778593131000,10482,{"hasNpmPackage":235,"license":247,"smitheryVerified":235},"Apache-2.0",{"hasCi":249,"hasTests":235},true,{"updatedAt":251},1778690836703,{"basePath":253,"githubOwner":222,"githubRepo":254,"locale":17,"slug":12,"type":255},"skills/huggingface-llm-trainer","skills","plugin",{"_creationTime":257,"_id":258,"community":259,"display":260,"identity":265,"parentExtension":268,"providers":269,"relations":285,"tags":287,"workflow":288},1778690773482.4824,"k17es3r8wd37t5rrwqcpp5kwrh86mxx8",{"reviewCount":8},{"description":261,"installMethods":262,"name":264,"sourceUrl":13},"Agent Skills for AI/ML tasks including dataset creation, model training, evaluation, and research paper publishing on Hugging Face Hub",{"claudeCode":263},"huggingface/skills","huggingface-skills",{"basePath":266,"githubOwner":222,"githubRepo":254,"locale":17,"slug":254,"type":267},"","marketplace",null,{"evaluate":270,"extract":279},{"promptVersionExtension":271,"promptVersionScoring":213,"score":272,"tags":273,"targetMarket":226,"tier":227},"3.1.0",95,[274,222,275,276,277,278],"ai-ml","datasets","models","research","developer-tools",{"commitSha":280,"marketplace":281,"plugin":283},"HEAD",{"name":264,"pluginCount":282},14,{"mcpCount":8,"provider":284,"skillCount":8},"classify",{"repoId":286},"kd72xwt5xnc0ktc4p7smzfcp3986m959",[274,275,278,222,276,277],{"evaluatedAt":289,"extractAt":290,"updatedAt":289},1778690814090,1778690773482,{"evaluate":292,"extract":294},{"promptVersionExtension":212,"promptVersionScoring":213,"score":216,"tags":293,"targetMarket":226,"tier":227},[219,220,221,222,223,224,225],{"commitSha":280},{"parentExtensionId":258,"repoId":286},{"_creationTime":297,"_id":286,"identity":298,"providers":299,"workflow":738},1778689536128.5474,{"githubOwner":222,"githubRepo":254,"sourceUrl":13},{"classify":300,"discover":731,"github":734},{"commitSha":280,"extensions":301},[302,315,321,329,337,345,353,361,369,377,385,393,401,409,417,425,468,477,483,489,506,512,519,561,572,591,597,617,629,653,711],{"basePath":266,"description":261,"displayName":264,"installMethods":303,"rationale":304,"selectedPaths":305,"source":314,"sourceLanguage":17,"type":267},{"claudeCode":263},"marketplace.json at .claude-plugin/marketplace.json",[306,309,311],{"path":307,"priority":308},".claude-plugin/marketplace.json","mandatory",{"path":310,"priority":308},"README.md",{"path":312,"priority":313},"LICENSE","high","rule",{"basePath":253,"description":10,"displayName":12,"installMethods":316,"rationale":317,"selectedPaths":318,"source":314,"sourceLanguage":17,"type":255},{"claudeCode":12},"inline plugin source from marketplace.json at skills/huggingface-llm-trainer",[319],{"path":320,"priority":313},"SKILL.md",{"basePath":322,"description":323,"displayName":324,"installMethods":325,"rationale":326,"selectedPaths":327,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-local-models","Use 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-local-models",{"claudeCode":324},"inline plugin source from marketplace.json at skills/huggingface-local-models",[328],{"path":320,"priority":313},{"basePath":330,"description":331,"displayName":332,"installMethods":333,"rationale":334,"selectedPaths":335,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-paper-publisher","Publish and manage research papers on Hugging Face Hub. Supports creating paper pages, linking papers to models/datasets, claiming authorship, and generating professional markdown-based research articles.","huggingface-paper-publisher",{"claudeCode":332},"inline plugin source from marketplace.json at skills/huggingface-paper-publisher",[336],{"path":320,"priority":313},{"basePath":338,"description":339,"displayName":340,"installMethods":341,"rationale":342,"selectedPaths":343,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-papers","Look 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-papers",{"claudeCode":340},"inline plugin source from marketplace.json at skills/huggingface-papers",[344],{"path":320,"priority":313},{"basePath":346,"description":347,"displayName":348,"installMethods":349,"rationale":350,"selectedPaths":351,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-community-evals","Add and manage evaluation results in Hugging Face model cards. Supports extracting eval tables from README content, importing scores from Artificial Analysis API, and running custom evaluations with vLLM/lighteval.","huggingface-community-evals",{"claudeCode":348},"inline plugin source from marketplace.json at skills/huggingface-community-evals",[352],{"path":320,"priority":313},{"basePath":354,"description":355,"displayName":356,"installMethods":357,"rationale":358,"selectedPaths":359,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-best","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.","huggingface-best",{"claudeCode":356},"inline plugin source from marketplace.json at skills/huggingface-best",[360],{"path":320,"priority":313},{"basePath":362,"description":363,"displayName":364,"installMethods":365,"rationale":366,"selectedPaths":367,"source":314,"sourceLanguage":17,"type":255},"skills/hf-cli","Execute Hugging Face Hub operations using the hf CLI. Download models/datasets, upload files, manage repos, and run cloud compute jobs.","hf-cli",{"claudeCode":364},"inline plugin source from marketplace.json at skills/hf-cli",[368],{"path":320,"priority":313},{"basePath":370,"description":371,"displayName":372,"installMethods":373,"rationale":374,"selectedPaths":375,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-trackio","Track 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.","huggingface-trackio",{"claudeCode":372},"inline plugin source from marketplace.json at skills/huggingface-trackio",[376],{"path":320,"priority":313},{"basePath":378,"description":379,"displayName":380,"installMethods":381,"rationale":382,"selectedPaths":383,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-datasets","Explore, query, and extract data from any Hugging Face dataset using the Dataset Viewer REST API and npx tooling. Zero Python dependencies — covers split/config discovery, row pagination, text search, filtering, SQL via parquetlens, and dataset upload via CLI.","huggingface-datasets",{"claudeCode":380},"inline plugin source from marketplace.json at skills/huggingface-datasets",[384],{"path":320,"priority":313},{"basePath":386,"description":387,"displayName":388,"installMethods":389,"rationale":390,"selectedPaths":391,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-tool-builder","Build reusable scripts for Hugging Face Hub and API workflows. Useful for chaining API calls, enriching Hub metadata, or automating repeated tasks.","huggingface-tool-builder",{"claudeCode":388},"inline plugin source from marketplace.json at skills/huggingface-tool-builder",[392],{"path":320,"priority":313},{"basePath":394,"description":395,"displayName":396,"installMethods":397,"rationale":398,"selectedPaths":399,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-gradio","Build Gradio web UIs and demos in Python. Use when creating or editing Gradio apps, components, event listeners, layouts, or chatbots.","huggingface-gradio",{"claudeCode":396},"inline plugin source from marketplace.json at skills/huggingface-gradio",[400],{"path":320,"priority":313},{"basePath":402,"description":403,"displayName":404,"installMethods":405,"rationale":406,"selectedPaths":407,"source":314,"sourceLanguage":17,"type":255},"skills/transformers-js","Run state-of-the-art machine learning models directly in JavaScript/TypeScript for NLP, computer vision, audio processing, and multimodal tasks. Works in Node.js and browsers with WebGPU/WASM using Hugging Face models.","transformers-js",{"claudeCode":404},"inline plugin source from marketplace.json at skills/transformers-js",[408],{"path":320,"priority":313},{"basePath":410,"description":411,"displayName":412,"installMethods":413,"rationale":414,"selectedPaths":415,"source":314,"sourceLanguage":17,"type":255},"skills/huggingface-vision-trainer","Train and fine-tune object detection models (RTDETRv2, YOLOS, DETR and others) and image classification models (timm and transformers models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3) using Transformers Trainer API on Hugging Face Jobs infrastructure or locally. Includes COCO dataset format support, Albumentations augmentation, mAP/mAR metrics, trackio tracking, hardware selection, and Hub persistence.","huggingface-vision-trainer",{"claudeCode":412},"inline plugin source from marketplace.json at skills/huggingface-vision-trainer",[416],{"path":320,"priority":313},{"basePath":418,"description":419,"displayName":420,"installMethods":421,"rationale":422,"selectedPaths":423,"source":314,"sourceLanguage":17,"type":255},"skills/train-sentence-transformers","Train or fine-tune sentence-transformers models across all three architectures: SentenceTransformer (bi-encoder embeddings), CrossEncoder (rerankers), and SparseEncoder (SPLADE). Covers loss selection, hard-negative mining, evaluators, distillation, LoRA, Matryoshka, and Hugging Face Hub publishing.","train-sentence-transformers",{"claudeCode":420},"inline plugin source from marketplace.json at skills/train-sentence-transformers",[424],{"path":320,"priority":313},{"basePath":266,"description":261,"displayName":264,"installMethods":426,"license":247,"rationale":427,"selectedPaths":428,"source":314,"sourceLanguage":17,"type":255},{"claudeCode":264},"plugin manifest at .claude-plugin/plugin.json",[429,431,432,433,436,438,440,442,444,446,448,450,452,454,456,458,460,462,464,466],{"path":430,"priority":308},".claude-plugin/plugin.json",{"path":310,"priority":308},{"path":312,"priority":313},{"path":434,"priority":435},"skills/hf-cli/SKILL.md","medium",{"path":437,"priority":435},"skills/huggingface-best/SKILL.md",{"path":439,"priority":435},"skills/huggingface-community-evals/SKILL.md",{"path":441,"priority":435},"skills/huggingface-datasets/SKILL.md",{"path":443,"priority":435},"skills/huggingface-gradio/SKILL.md",{"path":445,"priority":435},"skills/huggingface-llm-trainer/SKILL.md",{"path":447,"priority":435},"skills/huggingface-local-models/SKILL.md",{"path":449,"priority":435},"skills/huggingface-paper-publisher/SKILL.md",{"path":451,"priority":435},"skills/huggingface-papers/SKILL.md",{"path":453,"priority":435},"skills/huggingface-tool-builder/SKILL.md",{"path":455,"priority":435},"skills/huggingface-trackio/SKILL.md",{"path":457,"priority":435},"skills/huggingface-vision-trainer/SKILL.md",{"path":459,"priority":435},"skills/train-sentence-transformers/SKILL.md",{"path":461,"priority":435},"skills/transformers-js/SKILL.md",{"path":463,"priority":308},".mcp.json",{"path":465,"priority":313},"agents/AGENTS.md",{"path":467,"priority":313},".cursor-plugin/plugin.json",{"basePath":469,"description":470,"displayName":471,"installMethods":472,"rationale":473,"selectedPaths":474,"source":314,"sourceLanguage":17,"type":476},"hf-mcp/skills/hf-mcp","Use Hugging Face Hub via MCP server tools. Search models, datasets, Spaces, papers. Get repo details, fetch documentation, run compute jobs, and use Gradio Spaces as AI tools. Available when connected to the HF MCP server.","hf-mcp",{"claudeCode":263},"SKILL.md frontmatter at hf-mcp/skills/hf-mcp/SKILL.md",[475],{"path":320,"priority":308},"skill",{"basePath":362,"description":478,"displayName":364,"installMethods":479,"rationale":480,"selectedPaths":481,"source":314,"sourceLanguage":17,"type":476},"Hugging Face Hub CLI (`hf`) for downloading, uploading, and managing models, datasets, spaces, buckets, repos, papers, jobs, and more on the Hugging Face Hub. Use when: handling authentication; managing local cache; managing Hugging Face Buckets; running or scheduling jobs on Hugging Face infrastructure; managing Hugging Face repos; discussions and pull requests; browsing models, datasets and spaces; reading, searching, or browsing academic papers; managing collections; querying datasets; configuring spaces; setting up webhooks; or deploying and managing HF Inference Endpoints. Make sure to use this skill whenever the user mentions 'hf', 'huggingface', 'Hugging Face', 'huggingface-cli', or 'hugging face cli', or wants to do anything related to the Hugging Face ecosystem and to AI and ML in general. Also use for cloud storage needs like training checkpoints, data pipelines, or agent traces. Use even if the user doesn't explicitly ask for a CLI command. Replaces the deprecated `huggingface-cli`.",{"claudeCode":263},"SKILL.md frontmatter at skills/hf-cli/SKILL.md",[482],{"path":320,"priority":308},{"basePath":354,"description":484,"displayName":356,"installMethods":485,"rationale":486,"selectedPaths":487,"source":314,"sourceLanguage":17,"type":476},"Use when the user asks about finding the best, top, or recommended model for a task, wants to know what AI model to use, or wants to compare models by benchmark scores. Triggers on: \"best model for X\", \"what model should I use for\", \"top models for [task]\", \"which model runs on my laptop/machine/device\", \"recommend a model for\", \"what LLM should I use for\", \"compare models for\", \"what's state of the art for\", or any question about choosing an AI model for a specific use case. Always use this skill when the user wants model recommendations or comparisons, even if they don't explicitly mention HuggingFace or benchmarks.\n",{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-best/SKILL.md",[488],{"path":320,"priority":308},{"basePath":346,"description":490,"displayName":348,"installMethods":491,"rationale":492,"selectedPaths":493,"source":314,"sourceLanguage":17,"type":476},"Run evaluations for Hugging Face Hub models using inspect-ai and lighteval on local hardware. Use for backend selection, local GPU evals, and choosing between vLLM / Transformers / accelerate. Not for HF Jobs orchestration, model-card PRs, .eval_results publication, or community-evals automation.",{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-community-evals/SKILL.md",[494,495,498,500,502,504],{"path":320,"priority":308},{"path":496,"priority":497},"examples/.env.example","low",{"path":499,"priority":497},"examples/USAGE_EXAMPLES.md",{"path":501,"priority":497},"scripts/inspect_eval_uv.py",{"path":503,"priority":497},"scripts/inspect_vllm_uv.py",{"path":505,"priority":497},"scripts/lighteval_vllm_uv.py",{"basePath":378,"description":507,"displayName":380,"installMethods":508,"rationale":509,"selectedPaths":510,"source":314,"sourceLanguage":17,"type":476},"Use this skill for Hugging Face Dataset Viewer API workflows that fetch subset/split metadata, paginate rows, search text, apply filters, download parquet URLs, and read size or statistics.\r",{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-datasets/SKILL.md",[511],{"path":320,"priority":308},{"basePath":394,"description":395,"displayName":396,"installMethods":513,"rationale":514,"selectedPaths":515,"source":314,"sourceLanguage":17,"type":476},{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-gradio/SKILL.md",[516,517],{"path":320,"priority":308},{"path":518,"priority":435},"examples.md",{"basePath":253,"description":520,"displayName":12,"installMethods":521,"rationale":522,"selectedPaths":523,"source":314,"sourceLanguage":17,"type":476},"Train or fine-tune language and vision models using TRL (Transformer Reinforcement Learning) or Unsloth with Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV scripts with PEP 723 format, dataset preparation and validation, hardware selection, cost estimation, Trackio monitoring, Hub authentication, model selection/leaderboards and model persistence. Use for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup.",{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-llm-trainer/SKILL.md",[524,525,527,529,531,533,535,537,539,541,543,545,547,549,551,553,555,557,559],{"path":320,"priority":308},{"path":526,"priority":435},"references/gguf_conversion.md",{"path":528,"priority":435},"references/hardware_guide.md",{"path":530,"priority":435},"references/hub_saving.md",{"path":532,"priority":435},"references/local_training_macos.md",{"path":534,"priority":435},"references/reliability_principles.md",{"path":536,"priority":435},"references/trackio_guide.md",{"path":538,"priority":435},"references/training_methods.md",{"path":540,"priority":435},"references/training_patterns.md",{"path":542,"priority":435},"references/troubleshooting.md",{"path":544,"priority":435},"references/unsloth.md",{"path":546,"priority":497},"scripts/convert_to_gguf.py",{"path":548,"priority":497},"scripts/dataset_inspector.py",{"path":550,"priority":497},"scripts/estimate_cost.py",{"path":552,"priority":497},"scripts/hf_benchmarks.py",{"path":554,"priority":497},"scripts/train_dpo_example.py",{"path":556,"priority":497},"scripts/train_grpo_example.py",{"path":558,"priority":497},"scripts/train_sft_example.py",{"path":560,"priority":497},"scripts/unsloth_sft_example.py",{"basePath":322,"description":323,"displayName":324,"installMethods":562,"rationale":563,"selectedPaths":564,"source":314,"sourceLanguage":17,"type":476},{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-local-models/SKILL.md",[565,566,568,570],{"path":320,"priority":308},{"path":567,"priority":435},"references/hardware.md",{"path":569,"priority":435},"references/hub-discovery.md",{"path":571,"priority":435},"references/quantization.md",{"basePath":330,"description":331,"displayName":332,"installMethods":573,"rationale":574,"selectedPaths":575,"source":314,"sourceLanguage":17,"type":476},{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-paper-publisher/SKILL.md",[576,577,579,581,583,585,587,589],{"path":320,"priority":308},{"path":578,"priority":497},"examples/example_usage.md",{"path":580,"priority":435},"references/quick_reference.md",{"path":582,"priority":497},"scripts/paper_manager.py",{"path":584,"priority":497},"templates/arxiv.md",{"path":586,"priority":497},"templates/ml-report.md",{"path":588,"priority":497},"templates/modern.md",{"path":590,"priority":497},"templates/standard.md",{"basePath":338,"description":592,"displayName":340,"installMethods":593,"rationale":594,"selectedPaths":595,"source":314,"sourceLanguage":17,"type":476},"Look up and read Hugging Face paper pages in markdown, and use the papers API for structured metadata such as authors, linked models/datasets/spaces, Github repo and project page. Use when the user shares a Hugging Face paper page URL, an arXiv URL or ID, or asks to summarize, explain, or analyze an AI research paper.",{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-papers/SKILL.md",[596],{"path":320,"priority":308},{"basePath":386,"description":598,"displayName":388,"installMethods":599,"rationale":600,"selectedPaths":601,"source":314,"sourceLanguage":17,"type":476},"Use this skill when the user wants to build tool/scripts or achieve a task where using data from the Hugging Face API would help. This is especially useful when chaining or combining API calls or the task will be repeated/automated. This Skill creates a reusable script to fetch, enrich or process data.",{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-tool-builder/SKILL.md",[602,603,605,607,609,611,613,615],{"path":320,"priority":308},{"path":604,"priority":435},"references/baseline_hf_api.py",{"path":606,"priority":435},"references/baseline_hf_api.sh",{"path":608,"priority":435},"references/baseline_hf_api.tsx",{"path":610,"priority":435},"references/find_models_by_paper.sh",{"path":612,"priority":435},"references/hf_enrich_models.sh",{"path":614,"priority":435},"references/hf_model_card_frontmatter.sh",{"path":616,"priority":435},"references/hf_model_papers_auth.sh",{"basePath":370,"description":618,"displayName":372,"installMethods":619,"rationale":620,"selectedPaths":621,"source":314,"sourceLanguage":17,"type":476},"Track and visualize ML training experiments with Trackio. Use when logging metrics during training (Python API), firing alerts for training diagnostics, or retrieving/analyzing logged metrics (CLI). Supports real-time dashboard visualization, alerts with webhooks, HF Space syncing, and JSON output for automation.",{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-trackio/SKILL.md",[622,623,625,627],{"path":320,"priority":308},{"path":624,"priority":435},"references/alerts.md",{"path":626,"priority":435},"references/logging_metrics.md",{"path":628,"priority":435},"references/retrieving_metrics.md",{"basePath":410,"description":630,"displayName":412,"installMethods":631,"rationale":632,"selectedPaths":633,"source":314,"sourceLanguage":17,"type":476},"Trains and fine-tunes vision models for object detection (D-FINE, RT-DETR v2, DETR, YOLOS), image classification (timm models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3 — plus any Transformers classifier), and SAM/SAM2 segmentation using Hugging Face Transformers on Hugging Face Jobs cloud GPUs. Covers COCO-format dataset preparation, Albumentations augmentation, mAP/mAR evaluation, accuracy metrics, SAM segmentation with bbox/point prompts, DiceCE loss, hardware selection, cost estimation, Trackio monitoring, and Hub persistence. Use when users mention training object detection, image classification, SAM, SAM2, segmentation, image matting, DETR, D-FINE, RT-DETR, ViT, timm, MobileNet, ResNet, bounding box models, or fine-tuning vision models on Hugging Face Jobs.",{"claudeCode":263},"SKILL.md frontmatter at skills/huggingface-vision-trainer/SKILL.md",[634,635,637,638,640,642,643,645,646,647,649,651],{"path":320,"priority":308},{"path":636,"priority":435},"references/finetune_sam2_trainer.md",{"path":530,"priority":435},{"path":639,"priority":435},"references/image_classification_training_notebook.md",{"path":641,"priority":435},"references/object_detection_training_notebook.md",{"path":534,"priority":435},{"path":644,"priority":435},"references/timm_trainer.md",{"path":548,"priority":497},{"path":550,"priority":497},{"path":648,"priority":497},"scripts/image_classification_training.py",{"path":650,"priority":497},"scripts/object_detection_training.py",{"path":652,"priority":497},"scripts/sam_segmentation_training.py",{"basePath":418,"description":654,"displayName":420,"installMethods":655,"rationale":656,"selectedPaths":657,"source":314,"sourceLanguage":17,"type":476},"Train or fine-tune sentence-transformers models across `SentenceTransformer` (bi-encoder; dense or static embedding model; for retrieval, similarity, clustering, classification, paraphrase mining, dedup, multimodal), `CrossEncoder` (reranker; pair scoring for two-stage retrieval / pair classification), and `SparseEncoder` (SPLADE, sparse embedding model; for learned-sparse retrieval). Covers loss selection, hard-negative mining, evaluators, distillation, LoRA, Matryoshka, and Hugging Face Hub publishing. Use for any sentence-transformers training task.",{"claudeCode":263},"SKILL.md frontmatter at skills/train-sentence-transformers/SKILL.md",[658,659,661,663,665,667,669,670,672,674,676,678,680,682,684,685,687,689,691,693,695,697,699,701,703,705,707,709],{"path":320,"priority":308},{"path":660,"priority":435},"references/base_model_selection.md",{"path":662,"priority":435},"references/dataset_formats.md",{"path":664,"priority":435},"references/evaluators_cross_encoder.md",{"path":666,"priority":435},"references/evaluators_sentence_transformer.md",{"path":668,"priority":435},"references/evaluators_sparse_encoder.md",{"path":528,"priority":435},{"path":671,"priority":435},"references/hf_jobs_execution.md",{"path":673,"priority":435},"references/losses_cross_encoder.md",{"path":675,"priority":435},"references/losses_sentence_transformer.md",{"path":677,"priority":435},"references/losses_sparse_encoder.md",{"path":679,"priority":435},"references/model_architectures.md",{"path":681,"priority":435},"references/prompts_and_instructions.md",{"path":683,"priority":435},"references/training_args.md",{"path":542,"priority":435},{"path":686,"priority":497},"scripts/mine_hard_negatives.py",{"path":688,"priority":497},"scripts/train_cross_encoder_distillation_example.py",{"path":690,"priority":497},"scripts/train_cross_encoder_example.py",{"path":692,"priority":497},"scripts/train_cross_encoder_listwise_example.py",{"path":694,"priority":497},"scripts/train_sentence_transformer_distillation_example.py",{"path":696,"priority":497},"scripts/train_sentence_transformer_example.py",{"path":698,"priority":497},"scripts/train_sentence_transformer_make_multilingual_example.py",{"path":700,"priority":497},"scripts/train_sentence_transformer_matryoshka_example.py",{"path":702,"priority":497},"scripts/train_sentence_transformer_multi_dataset_example.py",{"path":704,"priority":497},"scripts/train_sentence_transformer_static_embedding_example.py",{"path":706,"priority":497},"scripts/train_sentence_transformer_with_lora_example.py",{"path":708,"priority":497},"scripts/train_sparse_encoder_distillation_example.py",{"path":710,"priority":497},"scripts/train_sparse_encoder_example.py",{"basePath":402,"description":712,"displayName":404,"installMethods":713,"rationale":714,"selectedPaths":715,"source":314,"sourceLanguage":17,"type":476},"Use Transformers.js to run state-of-the-art machine learning models directly in JavaScript/TypeScript. Supports NLP (text classification, translation, summarization), computer vision (image classification, object detection), audio (speech recognition, audio classification), and multimodal tasks. Works in browsers and server-side runtimes (Node.js, Bun, Deno) with WebGPU/WASM using pre-trained models from Hugging Face Hub.",{"claudeCode":263},"SKILL.md frontmatter at skills/transformers-js/SKILL.md",[716,717,719,721,723,725,727,729],{"path":320,"priority":308},{"path":718,"priority":435},"references/CACHE.md",{"path":720,"priority":435},"references/CONFIGURATION.md",{"path":722,"priority":435},"references/EXAMPLES.md",{"path":724,"priority":435},"references/MODEL_ARCHITECTURES.md",{"path":726,"priority":435},"references/MODEL_REGISTRY.md",{"path":728,"priority":435},"references/PIPELINE_OPTIONS.md",{"path":730,"priority":435},"references/TEXT_GENERATION.md",{"sources":732},[733],"manual",{"closedIssues90d":241,"description":735,"forks":242,"homepage":736,"license":247,"openIssues90d":243,"pushedAt":244,"readmeSize":239,"stars":245,"topics":737},"Give your agents the power of the Hugging Face ecosystem","https://huggingface.co",[],{"classifiedAt":739,"discoverAt":740,"extractAt":741,"githubAt":741,"updatedAt":739},1778690772996,1778689536128,1778690770714,[221,224,222,219,225,220,223],{"evaluatedAt":251,"extractAt":290,"updatedAt":251},[],[746,764,783,798,819,850],{"_creationTime":747,"_id":748,"community":749,"display":750,"identity":752,"providers":753,"relations":760,"tags":761,"workflow":762},1778690773482.483,"k172w5kkdn7117xpqyyhcqsww186n17b",{"reviewCount":8},{"description":323,"installMethods":751,"name":324,"sourceUrl":13},{"claudeCode":324},{"basePath":322,"githubOwner":222,"githubRepo":254,"locale":17,"slug":324,"type":255},{"evaluate":754,"extract":759},{"promptVersionExtension":212,"promptVersionScoring":213,"score":216,"tags":755,"targetMarket":226,"tier":227},[219,756,757,224,222,758],"local-models","llama-cpp","ml-ops",{"commitSha":280},{"parentExtensionId":258,"repoId":286},[224,222,757,219,756,758],{"evaluatedAt":763,"extractAt":290,"updatedAt":763},1778690852474,{"_creationTime":765,"_id":766,"community":767,"display":768,"identity":770,"providers":771,"relations":779,"tags":780,"workflow":781},1778690773482.4844,"k1785s3263hzxs6ne1nsbzas5186m86v",{"reviewCount":8},{"description":371,"installMethods":769,"name":372,"sourceUrl":13},{"claudeCode":372},{"basePath":370,"githubOwner":222,"githubRepo":254,"locale":17,"slug":372,"type":255},{"evaluate":772,"extract":778},{"promptVersionExtension":212,"promptVersionScoring":213,"score":216,"tags":773,"targetMarket":226,"tier":227},[225,774,775,222,776,777],"experiment-tracking","monitoring","python","cli",{"commitSha":280},{"parentExtensionId":258,"repoId":286},[777,774,222,225,775,776],{"evaluatedAt":782,"extractAt":290,"updatedAt":782},1778690988981,{"_creationTime":784,"_id":785,"community":786,"display":787,"identity":789,"providers":790,"relations":794,"tags":795,"workflow":796},1778690773482.4841,"k172m7ench1yy64c6nbcx5pmm586mphh",{"reviewCount":8},{"description":363,"installMethods":788,"name":364,"sourceUrl":13},{"claudeCode":364},{"basePath":362,"githubOwner":222,"githubRepo":254,"locale":17,"slug":364,"type":255},{"evaluate":791,"extract":793},{"promptVersionExtension":212,"promptVersionScoring":213,"score":216,"tags":792,"targetMarket":226,"tier":227},[222,777,278,225,275,276],{"commitSha":280},{"parentExtensionId":258,"repoId":286},[777,275,278,222,225,276],{"evaluatedAt":797,"extractAt":290,"updatedAt":797},1778690967596,{"_creationTime":799,"_id":800,"community":801,"display":802,"identity":805,"providers":806,"relations":815,"tags":816,"workflow":817},1778690773482.4834,"k179sm2kkyd7r7nz9jsx62jm9x86mw4a",{"reviewCount":8},{"description":339,"installMethods":803,"name":804,"sourceUrl":13},{"claudeCode":340},"Hugging Face Papers",{"basePath":338,"githubOwner":222,"githubRepo":254,"locale":17,"slug":340,"type":255},{"evaluate":807,"extract":814},{"promptVersionExtension":212,"promptVersionScoring":213,"score":808,"tags":809,"targetMarket":226,"tier":227},100,[222,810,811,812,277,813],"papers","arxiv","ai","metadata",{"commitSha":280,"license":247},{"parentExtensionId":258,"repoId":286},[812,811,222,813,810,277],{"evaluatedAt":818,"extractAt":290,"updatedAt":818},1778690901306,{"_creationTime":820,"_id":821,"community":822,"display":823,"identity":829,"providers":833,"relations":844,"tags":846,"workflow":847},1778675056600.209,"k1773g3pj6n54axnj39fwn0zfs86m1ay",{"reviewCount":8},{"description":824,"installMethods":825,"name":827,"sourceUrl":828},"Use when managing prompts in production at scale: versioning prompts, running A/B tests on prompts, building prompt registries, preventing prompt regressions, or creating eval pipelines for production",{"claudeCode":826},"prompt-governance","Prompt Governance","https://github.com/alirezarezvani/claude-skills",{"basePath":830,"githubOwner":831,"githubRepo":832,"locale":17,"slug":826,"type":255},"engineering/prompt-governance","alirezarezvani","claude-skills",{"evaluate":834,"extract":840},{"promptVersionExtension":212,"promptVersionScoring":213,"score":216,"tags":835,"targetMarket":226,"tier":227},[836,837,838,225,839],"prompt-engineering","prompt-management","production-ai","compliance",{"commitSha":280,"license":841,"plugin":842},"MIT",{"mcpCount":8,"provider":284,"skillCount":843},1,{"repoId":845},"kd7ff9s1w43mfyy1n7hf87816186m6px",[839,225,838,836,837],{"evaluatedAt":848,"extractAt":849,"updatedAt":848},1778676130492,1778675056600,{"_creationTime":851,"_id":852,"community":853,"display":854,"identity":857,"providers":858,"relations":865,"tags":866,"workflow":867},1778690773482.486,"k175g1spb5757qt4tnj9cktcn986mshy",{"reviewCount":8},{"description":261,"installMethods":855,"name":856,"sourceUrl":13},{"claudeCode":264},"Hugging Face Skills",{"basePath":266,"githubOwner":222,"githubRepo":254,"locale":17,"slug":254,"type":255},{"evaluate":859,"extract":863},{"promptVersionExtension":212,"promptVersionScoring":213,"score":860,"tags":861,"targetMarket":226,"tier":227},98,[222,812,862,275,276,220,777,776],"ml",{"commitSha":280,"license":247,"plugin":864},{"mcpCount":8,"provider":284,"skillCount":282},{"repoId":286},[812,777,275,222,862,276,776,220],{"evaluatedAt":868,"extractAt":290,"updatedAt":868},1778691185872]