[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7DH5STQZH232A2K7QBCVGQZ":3,"skill-same:01M7DH5STQZH232A2K7QBCVGQZ":44},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":13,"repositoryHtmlUrl":26,"repositoryHomepageUrl":27,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M7DH5STQZH232A2K7QBCVGQZ","huggingface-best",null,"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\u002Fmachine\u002Fdevice\", \"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.","ak2dev\u002Fagentguard","ak2dev","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F233005497?v=4","01M7DGWK3E3XCC276C6YRFSN05","01M7DGWK4J2A9WQ44CS5CRP4R6",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard\u002Fblob\u002Fmain\u002Fbench\u002Fcorpus\u002Fbenign\u002Freal-hf-huggingface-best\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: huggingface-best\ndescription: >-\n  Use when the user asks about finding the best, top, or recommended model for a task,\n  wants to know what AI model to use, or wants to compare models by benchmark scores.\n  Triggers on: \"best model for X\", \"what model should I use for\", \"top models for [task]\",\n  \"which model runs on my laptop\u002Fmachine\u002Fdevice\", \"recommend a model for\", \"what LLM should\n  I use for\", \"compare models for\", \"what's state of the art for\", or any question about\n  choosing an AI model for a specific use case. Always use this skill when the user wants\n  model recommendations or comparisons, even if they don't explicitly mention HuggingFace\n  or benchmarks.\n---\n\n# HuggingFace Best Model Finder\n\nFinds the best models for a task by querying official HF benchmark leaderboards, enriching\nresults with model size data, filtering for what fits on the user's device, and returning a\ncomparison table with benchmark scores.\n\n---\n\n## Step 1: Parse the request\n\nExtract from the user's message:\n- **Task**: what they want the model to do (coding, math\u002Freasoning, chat, OCR, RAG\u002Fretrieval, speech recognition, image classification, multimodal, agents, etc.)\n- **Device**: hardware constraints (MacBook M-series 8\u002F16\u002F32\u002F64GB unified memory, RTX GPU with VRAM amount, CPU-only, cloud\u002Fno constraint, etc.)\n\nIf device is not mentioned, skip filtering entirely and return the highest-performing models regardless of size. If the task is genuinely ambiguous, ask one clarifying question.\n\n### Device → max parameter budget\n\nWhen a device is specified, extract its available memory (unified RAM for Apple Silicon, VRAM for discrete GPUs) and apply:\n\n- **fp16 max params (B)** ≈ memory (GB) ÷ 2\n- **Q4 max params (B)** ≈ memory (GB) × 2\n\nExamples: 16GB → 8B fp16 \u002F 32B Q4 — 24GB VRAM → 12B fp16 \u002F 48B Q4 — 8GB → 4B fp16 \u002F 16B Q4\n\n---\n\n## Step 2: Find relevant benchmark datasets\n\nFetch the full list of official HF benchmarks:\n\n```bash\ncurl -s -H \"Authorization: Bearer $(cat ~\u002F.cache\u002Fhuggingface\u002Ftoken)\" \\\n  \"https:\u002F\u002Fhuggingface.co\u002Fapi\u002Fdatasets?filter=benchmark:official&limit=500\" | jq '[.[] | {id, tags, description}]'\n```\n\nRead the returned list and select the datasets most relevant to the user's task — match on dataset id, tags, and description. Use your judgment; don't limit yourself to 2-3. Aim for comprehensive coverage: if 5 benchmarks clearly cover the task, use all 5.\n\n---\n\n## Step 3: Fetch top models from leaderboards\n\nFor each selected benchmark dataset:\n\n```bash\ncurl -s -H \"Authorization: Bearer $(cat ~\u002F.cache\u002Fhuggingface\u002Ftoken)\" \\\n  \"https:\u002F\u002Fhuggingface.co\u002Fapi\u002Fdatasets\u002F\u003Cnamespace>\u002F\u003Crepo>\u002Fleaderboard\" | jq '[.[:15] | .[] | {rank, modelId, value, verified}]'\n```\n\nCollect model IDs and scores across all benchmarks. If a leaderboard returns an error (404, 401, etc.), skip it and note it in the output.\n\n---\n\n## Step 4: Enrich with model metadata\n\nFor the top 10-15 candidate model IDs, get model infos.\n\n```bash\n# REST API\ncurl -s -H \"Authorization: Bearer $(cat ~\u002F.cache\u002Fhuggingface\u002Ftoken)\" \\\n  \"https:\u002F\u002Fhuggingface.co\u002Fapi\u002Fmodels\u002Forg\u002Fmodel1\" | jq '{safetensors, tags, cardData}'\n\n# CLI (hf-cli)\nhf models info org\u002Fmodel1 --json | jq '{safetensors, tags, cardData}'\n```\n\nExtract from each response:\n- **Parameters**: `safetensors.total` → convert to B (e.g., 7_241_748_480 → \"7.2B\")\n- **License**: from model card tags (look for `license:apache-2.0`, `license:mit`, etc.)\n- If `safetensors` is absent, parse size from the model name (look for \"7b\", \"8b\", \"13b\", \"70b\", \"72b\", etc.)\n\n---\n\n## Step 5: Filter and rank\n\n**If a device was specified:**\n1. Remove models exceeding the fp16 parameter budget for the device\n2. Flag models that fit only with Q4 quantization (multiply budget by ~4 for Q4 capacity)\n3. If a highly-ranked model is slightly over budget, keep it with a \"needs Q4\" note — don't silently drop it\n\n**If no device was mentioned:** skip all size filtering — just rank by benchmark score.\n\nThen: rank by benchmark score (descending), keep top 5-8 models.\n\nInclude proprietary models (GPT-4, Claude, Gemini) if they appear on leaderboards, but flag them as \"API only \u002F not self-hostable\". If the user explicitly asked for local\u002Fopen models only, exclude them.\n\n---\n\n## Step 6: Output\n\n### Comparison table\n\n```markdown\n| # | Model | Params | [Benchmark 1] | [Benchmark 2] | License | On device |\n|---|-------|--------|--------------|--------------|---------|-----------|\n| ⭐1 | [org\u002Fname](https:\u002F\u002Fhuggingface.co\u002Forg\u002Fname) | 7B | 85.2% | — | Apache 2.0 | Yes (fp16) |\n| 2 | [org\u002Fname](https:\u002F\u002Fhuggingface.co\u002Forg\u002Fname) | 13B | 83.1% | 71.5% | MIT | Q4 only |\n| 3 | [org\u002Fname](https:\u002F\u002Fhuggingface.co\u002Forg\u002Fname) | 70B | 90.0% | 81.0% | Llama | Too large |\n```\n\n- Link model names to `https:\u002F\u002Fhuggingface.co\u002F\u003Cmodel_id>`\n- Use `—` for benchmarks where the model wasn't evaluated\n- Star the top recommended pick with ⭐\n- \"On device\" values: `Yes (fp16)`, `Q4 only`, `Too large`, `API only`\n\n### Follow-up\n\nAfter presenting the table, ask the user: \"Would you like to run **[top recommended model]**?\"\n\nIf they say yes, ask whether they'd prefer to:\n- **Run locally** — ask about their device if not already known, then give appropriate setup instructions\n- **Run on HF Jobs** — point them to the HF Jobs guide: https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Fhuggingface_hub\u002Fen\u002Fguides\u002Fjobs\n\n---\n\n## Error handling\n\n- **Leaderboard not found**: skip, note \"leaderboard unavailable\" in output\n- **Model missing from hub_repo_details**: fall back to parsing size from model name\n- **No benchmarks found for task**: use the curated fallback table above, or try `hub_repo_search` with `filters=[\"\u003Ctask>\"]` sorted by `trendingScore`\n- **All leaderboards fail**: fall back to `hub_repo_search` for popular models tagged with the task, note that results are by popularity rather than benchmark score\n","# HuggingFace Best Model Finder\n\nFinds the best models for a task by querying official HF benchmark leaderboards, enriching\nresults with model size data, filtering for what fits on the user's device, and returning a\ncomparison table with benchmark scores.\n\n---\n\n## Step 1: Parse the request\n\nExtract from the user's message:\n- **Task**: what they want the model to do (coding, math\u002Freasoning, chat, OCR, RAG\u002Fretrieval, speech recognition, image classification, multimodal, agents, etc.)\n- **Device**: hardware constraints (MacBook M-series 8\u002F16\u002F32\u002F64GB unified memory, RTX GPU with VRAM amount, CPU-only, cloud\u002Fno constraint, etc.)\n\nIf device is not mentioned, skip filtering entirely and return the highest-performing models regardless of size. If the task is genuinely ambiguous, ask one clarifying question.\n\n### Device → max parameter budget\n\nWhen a device is specified, extract its available memory (unified RAM for Apple Silicon, VRAM for discrete GPUs) and apply:\n\n- **fp16 max params (B)** ≈ memory (GB) ÷ 2\n- **Q4 max params (B)** ≈ memory (GB) × 2\n\nExamples: 16GB → 8B fp16 \u002F 32B Q4 — 24GB VRAM → 12B fp16 \u002F 48B Q4 — 8GB → 4B fp16 \u002F 16B Q4\n\n---\n\n## Step 2: Find relevant benchmark datasets\n\nFetch the full list of official HF benchmarks:\n\n```bash\ncurl -s -H \"Authorization: Bearer $(cat ~\u002F.cache\u002Fhuggingface\u002Ftoken)\" \\\n  \"https:\u002F\u002Fhuggingface.co\u002Fapi\u002Fdatasets?filter=benchmark:official&limit=500\" | jq '[.[] | {id, tags, description}]'\n```\n\nRead the returned list and select the datasets most relevant to the user's task — match on dataset id, tags, and description. Use your judgment; don't limit yourself to 2-3. Aim for comprehensive coverage: if 5 benchmarks clearly cover the task, use all 5.\n\n---\n\n## Step 3: Fetch top models from leaderboards\n\nFor each selected benchmark dataset:\n\n```bash\ncurl -s -H \"Authorization: Bearer $(cat ~\u002F.cache\u002Fhuggingface\u002Ftoken)\" \\\n  \"https:\u002F\u002Fhuggingface.co\u002Fapi\u002Fdatasets\u002F\u003Cnamespace>\u002F\u003Crepo>\u002Fleaderboard\" | jq '[.[:15] | .[] | {rank, modelId, value, verified}]'\n```\n\nCollect model IDs and scores across all benchmarks. If a leaderboard returns an error (404, 401, etc.), skip it and note it in the output.\n\n---\n\n## Step 4: Enrich with model metadata\n\nFor the top 10-15 candidate model IDs, get model infos.\n\n```bash\n# REST API\ncurl -s -H \"Authorization: Bearer $(cat ~\u002F.cache\u002Fhuggingface\u002Ftoken)\" \\\n  \"https:\u002F\u002Fhuggingface.co\u002Fapi\u002Fmodels\u002Forg\u002Fmodel1\" | jq '{safetensors, tags, cardData}'\n\n# CLI (hf-cli)\nhf models info org\u002Fmodel1 --json | jq '{safetensors, tags, cardData}'\n```\n\nExtract from each response:\n- **Parameters**: `safetensors.total` → convert to B (e.g., 7_241_748_480 → \"7.2B\")\n- **License**: from model card tags (look for `license:apache-2.0`, `license:mit`, etc.)\n- If `safetensors` is absent, parse size from the model name (look for \"7b\", \"8b\", \"13b\", \"70b\", \"72b\", etc.)\n\n---\n\n## Step 5: Filter and rank\n\n**If a device was specified:**\n1. Remove models exceeding the fp16 parameter budget for the device\n2. Flag models that fit only with Q4 quantization (multiply budget by ~4 for Q4 capacity)\n3. If a highly-ranked model is slightly over budget, keep it with a \"needs Q4\" note — don't silently drop it\n\n**If no device was mentioned:** skip all size filtering — just rank by benchmark score.\n\nThen: rank by benchmark score (descending), keep top 5-8 models.\n\nInclude proprietary models (GPT-4, Claude, Gemini) if they appear on leaderboards, but flag them as \"API only \u002F not self-hostable\". If the user explicitly asked for local\u002Fopen models only, exclude them.\n\n---\n\n## Step 6: Output\n\n### Comparison table\n\n```markdown\n| # | Model | Params | [Benchmark 1] | [Benchmark 2] | License | On device |\n|---|-------|--------|--------------|--------------|---------|-----------|\n| ⭐1 | [org\u002Fname](https:\u002F\u002Fhuggingface.co\u002Forg\u002Fname) | 7B | 85.2% | — | Apache 2.0 | Yes (fp16) |\n| 2 | [org\u002Fname](https:\u002F\u002Fhuggingface.co\u002Forg\u002Fname) | 13B | 83.1% | 71.5% | MIT | Q4 only |\n| 3 | [org\u002Fname](https:\u002F\u002Fhuggingface.co\u002Forg\u002Fname) | 70B | 90.0% | 81.0% | Llama | Too large |\n```\n\n- Link model names to `https:\u002F\u002Fhuggingface.co\u002F\u003Cmodel_id>`\n- Use `—` for benchmarks where the model wasn't evaluated\n- Star the top recommended pick with ⭐\n- \"On device\" values: `Yes (fp16)`, `Q4 only`, `Too large`, `API only`\n\n### Follow-up\n\nAfter presenting the table, ask the user: \"Would you like to run **[top recommended model]**?\"\n\nIf they say yes, ask whether they'd prefer to:\n- **Run locally** — ask about their device if not already known, then give appropriate setup instructions\n- **Run on HF Jobs** — point them to the HF Jobs guide: https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Fhuggingface_hub\u002Fen\u002Fguides\u002Fjobs\n\n---\n\n## Error handling\n\n- **Leaderboard not found**: skip, note \"leaderboard unavailable\" in output\n- **Model missing from hub_repo_details**: fall back to parsing size from model name\n- **No benchmarks found for task**: use the curated fallback table above, or try `hub_repo_search` with `filters=[\"\u003Ctask>\"]` sorted by `trendingScore`\n- **All leaderboards fail**: fall back to `hub_repo_search` for popular models tagged with the task, note that results are by popularity rather than benchmark score\n",{"name":5,"description":7},"main","634c25395a24b98cd6125453ad4abb65517bcabb","bench\u002Fcorpus\u002Fbenign\u002Freal-hf-huggingface-best","https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard","https:\u002F\u002Fsite-bionic8.vercel.app","https:\u002F\u002Fgithub.com\u002Fak2dev","2026-11-11 02:37:06",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add ak2dev\u002Fagentguard --skill huggingface-best",{"tool":36,"command":37},"bunx","bunx skills add ak2dev\u002Fagentguard --skill huggingface-best",{"tool":39,"command":40},"pnpm","pnpm dlx skills add ak2dev\u002Fagentguard --skill huggingface-best","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard\u002Fblob\u002Fmain\u002Fbench\u002Fcorpus\u002Fbenign\u002Freal-hf-huggingface-best\u002FSKILL.md",true,[],[45,52,57,62,67,73],{"skillNo":46,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":18},"01M7DJ7GBSJBJ32GXNQXVAMNYK","csv-tool","Summarize CSV files and report column statistics.","https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard\u002Fblob\u002Fmain\u002Ffixtures\u002Frules\u002FAG-SYS-002\u002Fpositive\u002Ffault\u002Fcsv-tool\u002FSKILL.md",[],[],{"skillNo":53,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":54,"categories":55,"keywords":56,"publishedAt":18},"01M7DJ7F00PA7EHFM39RVCC8R5","https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard\u002Fblob\u002Fmain\u002Ffixtures\u002Frules\u002FAG-SYS-002\u002Fnegative\u002Fnormal\u002Fcsv-tool\u002FSKILL.md",[],[],{"skillNo":58,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":59,"categories":60,"keywords":61,"publishedAt":18},"01M7DJ7DP7KDW7MYY5Q0R0JFEN","https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard\u002Fblob\u002Fmain\u002Ffixtures\u002Frules\u002FAG-SYS-001\u002Fpositive\u002Ftoo-large\u002Fcsv-tool\u002FSKILL.md",[],[],{"skillNo":63,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":18},"01M7DJ7C3VA0KQ8Q9ESNX6AJHE","https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard\u002Fblob\u002Fmain\u002Ffixtures\u002Frules\u002FAG-SYS-001\u002Fnegative\u002Fsmall\u002Fcsv-tool\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":18},"01M7DJ7AQZTGY5YADQX73TRB7K","pdf","https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard\u002Fblob\u002Fmain\u002Ffixtures\u002Frules\u002FAG-SKL-TYP-002\u002Fpositive\u002Fother-publisher\u002Fpdf\u002FSKILL.md",[],[],{"skillNo":74,"name":69,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":18},"01M7DJ7831XDVWS4W9SAAKBJHX","https:\u002F\u002Fgithub.com\u002Fak2dev\u002Fagentguard\u002Fblob\u002Fmain\u002Ffixtures\u002Frules\u002FAG-SKL-TYP-002\u002Fnegative\u002Fsame-publisher\u002Fpdf\u002FSKILL.md",[],[]]