[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EM9DRC9AJDMW2M2TGJFSBA":3,"skill-same:01M7EM9DRC9AJDMW2M2TGJFSBA":47},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":31,"ownerDescription":6,"updatedAt":32,"install":33,"resources":46},"01M7EM9DRC9AJDMW2M2TGJFSBA","ai21-guide",null,"Build with AI21 Labs' models — Jurassic for generation and strong task-specific models for enterprise use.","TanNhatCMS\u002Fjarvis","TanNhatCMS","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F113796420?v=4","01M753MV17KG5CD3152G3WQEQY","KietNT","01M7EKEDYVYQ6PWVJJMZ911VMX",0,"MIT","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fams-ai21-guide\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: ai21-guide\ndescription: Build with AI21 Labs' models — Jurassic for generation and strong task-specific models for enterprise use.\ncategory: ai-research\n---\n\n## Overview\n\nAI21 Labs builds foundation models with an enterprise focus: the Jurassic\nfamily for general generation and, distinctively, a set of task-specific models\n(Jamba for long-context, plus specialized models for summarization, reading\ncomprehension, and other focused tasks). The positioning is practical\nenterprise AI: strong performance on specific business tasks, long-context\ncapabilities, and managed APIs with enterprise controls.\n\nThe distinctive angle is task specialization: where most labs push one general\nmodel for everything, AI21 offers models tuned for particular jobs. For\nenterprises with well-defined high-volume tasks (summarization pipelines,\ndocument Q&A, classification), a specialized model can outperform a generalist\nat lower cost.\n\nEvaluate AI21 task by task: the question isn't \"is Jurassic the best general\nmodel\" but \"does the task-specific model beat the generalist on my task at my\neconomics.\"\n\n## When to use\n\n- High-volume summarization with consistent quality requirements.\n- Long-context tasks (Jamba's context windows suit document-scale work).\n- Document Q&A and reading comprehension at enterprise scale.\n- Task-specific models where specialization beats generalist prompting.\n- Enterprise deployments needing managed APIs with governance.\n- Evaluating alternatives to generalist frontier models on defined tasks.\n\n## Core concepts\n\n- **Jurassic models**: general-purpose generation models. Benchmark against\n  your current provider on your tasks — the comparison is task-specific.\n- **Jamba**: long-context models (hybrid architecture) for document-scale\n  inputs. Evaluate long-context quality on your documents, not just context\n  length claims.\n- **Task-specific models**: specialized models for summarization, comprehension,\n  and similar focused tasks. These are the differentiator — test them on the\n  tasks they're built for.\n- **Long-context handling**: large context windows for whole-document\n  processing. But length isn't quality — test retrieval and reasoning over\n  long inputs on your data.\n- **Enterprise API**: managed endpoints with the controls enterprises expect.\n  Evaluate the operational fit alongside model quality.\n- **Grounding**: models tuned for grounded enterprise use. Test faithfulness\n  on your RAG pipelines.\n- **Pricing**: per-token pricing; task-specific models may offer better\n  economics than generalists for their tasks. Model cost per completed task.\n- **Model selection**: the catalog has distinct models for distinct jobs —\n  choose deliberately per task, not one model for everything.\n\n## Practical workflow\n\n1. **Identify candidate tasks.** List your well-defined, high-volume tasks\n   (summarization, Q&A, classification). These are where task-specific models\n   may win.\n2. **Benchmark task-specific vs. generalist.** Same task, same eval set:\n   AI21's specialized model vs. your current generalist. Measure quality and\n   cost per task.\n3. **Test long-context honestly.** For Jamba: your longest real documents,\n   with questions requiring information from across the document. Test\n   \"lost in the middle\" behavior explicitly.\n4. **Evaluate grounding.** For RAG use: faithfulness to source documents,\n   measured on your corpus and queries.\n5. **Model the economics.** Task-specific models at lower per-token cost can\n   win big at volume — but only if quality clears your bar. Cost per completed\n   task, as always.\n6. **Pilot on production-shaped traffic.** Run a bounded pilot with real\n   inputs; monitor quality and latency distributions.\n7. **Decide per task.** The outcome may be hybrid: AI21 for summarization,\n   another provider for chat. That's fine — route per task.\n\nChecklist for AI21 in production:\n- Task-specific models benchmarked against generalists on your tasks.\n- Long-context quality tested on your real documents.\n- Grounding faithfulness measured for RAG use.\n- Cost per completed task modeled at volume.\n- Pilot validated on production-shaped traffic.\n\n## Common pitfalls\n\n- **Generalist comparisons only.** Evaluating Jurassic as a generic chatbot\n  and missing the task-specific models where AI21 actually differentiates.\n- **Context length as quality.** Assuming a big context window means good\n  long-document reasoning. Test with questions spanning the full input.\n- **No task-specific benchmarking.** Choosing models on vibes instead of\n  task-level evals. The specialization story needs task-level proof.\n- **Ignoring the hybrid outcome.** Feeling forced to pick one provider for\n  everything. Route per task — that's the rational architecture.\n- **Volume economics unmodeled.** Task-specific pricing advantages only matter\n  at volume. Model your actual throughput.\n- **Lost-in-the-middle blindness.** Long-context models still degrade on\n  information buried mid-document. Test explicitly.\n- **Single-task overfitting.** A model great at your summarization eval but\n  untested on your Q&A. Evaluate each deployed task separately.\n- **Skipping the pilot.** Going straight from benchmarks to full production.\n  Real traffic always differs — pilot first.\n","## Overview\n\nAI21 Labs builds foundation models with an enterprise focus: the Jurassic\nfamily for general generation and, distinctively, a set of task-specific models\n(Jamba for long-context, plus specialized models for summarization, reading\ncomprehension, and other focused tasks). The positioning is practical\nenterprise AI: strong performance on specific business tasks, long-context\ncapabilities, and managed APIs with enterprise controls.\n\nThe distinctive angle is task specialization: where most labs push one general\nmodel for everything, AI21 offers models tuned for particular jobs. For\nenterprises with well-defined high-volume tasks (summarization pipelines,\ndocument Q&A, classification), a specialized model can outperform a generalist\nat lower cost.\n\nEvaluate AI21 task by task: the question isn't \"is Jurassic the best general\nmodel\" but \"does the task-specific model beat the generalist on my task at my\neconomics.\"\n\n## When to use\n\n- High-volume summarization with consistent quality requirements.\n- Long-context tasks (Jamba's context windows suit document-scale work).\n- Document Q&A and reading comprehension at enterprise scale.\n- Task-specific models where specialization beats generalist prompting.\n- Enterprise deployments needing managed APIs with governance.\n- Evaluating alternatives to generalist frontier models on defined tasks.\n\n## Core concepts\n\n- **Jurassic models**: general-purpose generation models. Benchmark against\n  your current provider on your tasks — the comparison is task-specific.\n- **Jamba**: long-context models (hybrid architecture) for document-scale\n  inputs. Evaluate long-context quality on your documents, not just context\n  length claims.\n- **Task-specific models**: specialized models for summarization, comprehension,\n  and similar focused tasks. These are the differentiator — test them on the\n  tasks they're built for.\n- **Long-context handling**: large context windows for whole-document\n  processing. But length isn't quality — test retrieval and reasoning over\n  long inputs on your data.\n- **Enterprise API**: managed endpoints with the controls enterprises expect.\n  Evaluate the operational fit alongside model quality.\n- **Grounding**: models tuned for grounded enterprise use. Test faithfulness\n  on your RAG pipelines.\n- **Pricing**: per-token pricing; task-specific models may offer better\n  economics than generalists for their tasks. Model cost per completed task.\n- **Model selection**: the catalog has distinct models for distinct jobs —\n  choose deliberately per task, not one model for everything.\n\n## Practical workflow\n\n1. **Identify candidate tasks.** List your well-defined, high-volume tasks\n   (summarization, Q&A, classification). These are where task-specific models\n   may win.\n2. **Benchmark task-specific vs. generalist.** Same task, same eval set:\n   AI21's specialized model vs. your current generalist. Measure quality and\n   cost per task.\n3. **Test long-context honestly.** For Jamba: your longest real documents,\n   with questions requiring information from across the document. Test\n   \"lost in the middle\" behavior explicitly.\n4. **Evaluate grounding.** For RAG use: faithfulness to source documents,\n   measured on your corpus and queries.\n5. **Model the economics.** Task-specific models at lower per-token cost can\n   win big at volume — but only if quality clears your bar. Cost per completed\n   task, as always.\n6. **Pilot on production-shaped traffic.** Run a bounded pilot with real\n   inputs; monitor quality and latency distributions.\n7. **Decide per task.** The outcome may be hybrid: AI21 for summarization,\n   another provider for chat. That's fine — route per task.\n\nChecklist for AI21 in production:\n- Task-specific models benchmarked against generalists on your tasks.\n- Long-context quality tested on your real documents.\n- Grounding faithfulness measured for RAG use.\n- Cost per completed task modeled at volume.\n- Pilot validated on production-shaped traffic.\n\n## Common pitfalls\n\n- **Generalist comparisons only.** Evaluating Jurassic as a generic chatbot\n  and missing the task-specific models where AI21 actually differentiates.\n- **Context length as quality.** Assuming a big context window means good\n  long-document reasoning. Test with questions spanning the full input.\n- **No task-specific benchmarking.** Choosing models on vibes instead of\n  task-level evals. The specialization story needs task-level proof.\n- **Ignoring the hybrid outcome.** Feeling forced to pick one provider for\n  everything. Route per task — that's the rational architecture.\n- **Volume economics unmodeled.** Task-specific pricing advantages only matter\n  at volume. Model your actual throughput.\n- **Lost-in-the-middle blindness.** Long-context models still degrade on\n  information buried mid-document. Test explicitly.\n- **Single-task overfitting.** A model great at your summarization eval but\n  untested on your Q&A. Evaluate each deployed task separately.\n- **Skipping the pilot.** Going straight from benchmarks to full production.\n  Real traffic always differs — pilot first.\n",{"name":5,"category":24,"description":7},"ai-research","main","af388a77c2fe5e0b4fd879cca4300a0ffa9a6579","skills\u002Fams-ai21-guide",1,"https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis","https:\u002F\u002Fgithub.com\u002FTanNhatCMS","https:\u002F\u002Fmrkiet.name.vn","2026-11-07 14:23:08",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add TanNhatCMS\u002Fjarvis --skill ai21-guide",{"tool":39,"command":40},"bunx","bunx skills add TanNhatCMS\u002Fjarvis --skill ai21-guide",{"tool":42,"command":43},"pnpm","pnpm dlx skills add TanNhatCMS\u002Fjarvis --skill ai21-guide","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fams-ai21-guide\u002FSKILL.md",true,[],[48,55,63,70,77,84],{"skillNo":49,"name":50,"nameZh":6,"description":51,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":52,"categories":53,"keywords":54,"publishedAt":19},"01M7EWYJR0J8DARY7E59Z6Y9MV","web-scrape","Turn web pages into LLM-ready markdown via the Firecrawl API (renders JS, anti-bot handling). Use when browser_open\u002Ffetch fails on JS-heavy pages, or when you need clean markdown\u002Fstructured extraction from a site. Requires FIRECRAWL_API_KEY.","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fweb-scrape\u002FSKILL.md",[],[],{"skillNo":56,"name":57,"nameZh":6,"description":58,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":59,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M7EWYBN5P6F2AJG3FR24T9BP","design-taste-frontend","tasteskill v2, the anti-slop frontend skill. Art direction for websites, distilled from a study of 247 curated gallery sites (One Page Love, Awwwards Sites of the Day). Deals researched directions (idea, composition, type, palette, photographs, section blueprints) from a library, enforces a strict no-slop house style with a real-browser inspector, and runs a look-and-fix loop on screenshots. Use this whenever you build, design, redesign, restyle, polish or review a landing page, marketing site, portfolio, product page, one-pager, hero or any public-facing web page, in any stack (HTML\u002FCSS, React, Next.js, Vue, Svelte, Astro, Tailwind) — also when the user says a page looks generic, templated, \"AI-made\", sloppy, or asks to make it look designed, premium, award-level or less like every other site. Not for dashboards or app screens.","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Ftaste-design\u002FLICENSE","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Ftaste-design\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M7EWY8FQFNQ2GB7NZZWGSSDX","test-driven-development","Use when implementing any feature or bugfix, before writing implementation code","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-tdd\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7EWY5SHBFRH1HDHVWM9K7TV","systematic-debugging","Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-debugging\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7EWY3NM0Z0JAKS7HDZ24K66","requesting-code-review","Use when completing tasks, implementing major features, or before merging to verify work meets requirements","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-code-review\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M7EWY0WRMWX4M9ZFY0WKC593","brainstorming","You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).","https:\u002F\u002Fgithub.com\u002FTanNhatCMS\u002Fjarvis\u002Fblob\u002Fmain\u002Fskills\u002Fsuperpowers-brainstorming\u002FSKILL.md",[],[]]