# MYRILUM > Bilingual AI-native full-stack delivery platform, regional AI Skills Manifold index (CN build, snapshot 2026-11-21, 36 benchmark-indexed canonical skills), and regional AI Router Manifold index (40 canonical models). Executive Digest (EN): AI Skills Leaderboard (CN) top 3: find-skills / deepseek-r1-reasoning-distiller / myrilum-case-brief. Peak lift: skill-telemetry-probe (28.0%/1k tok). Fork alert: frontend-design-mega-bundle flags bloat; myrilum-geo-citation-auditor is a high-lift sleeper. Executive Digest (ZH): AI 技能 排行榜(CN)前三:find-skills / deepseek-r1-reasoning-distiller / myrilum-case-brief;最高单位增益 skill-telemetry-probe(28.0%/千词元)。分叉预警:frontend-design-mega-bundle 属上下文膨胀,myrilum-geo-citation-auditor 属高增益黑马。 Router Digest (EN): AI Router Manifold Index (CN ref baseline) top 3: deepseek/deepseek-v4-1-flash / minimax/minimax-m3-1-flash / zhipu/glm-5-3-flash. Peak value: openai/gpt-6-luna (507.8/$). Fork alert: openai/gpt-6-2-astra flags ref latency degradation; minimax/minimax-m2-flash is a high-value sleeper (unmeasured baseline). Router Digest (ZH): AI 模型路由流形总榜(CN 参考基线)前三:deepseek/deepseek-v4-1-flash / minimax/minimax-m3-1-flash / zhipu/glm-5-3-flash;峰值标价性价比 openai/gpt-6-luna(507.8/$)。分叉预警:openai/gpt-6-2-astra 属高基准参考延迟退化,minimax/minimax-m2-flash 属高性价比黑马(非实时遥测)。 ## Core Official Surfaces - [Home](https://myrilum.cn/): MYRILUM bilingual official overview - [Products & Services](https://myrilum.cn/market): Curated non-transactional capability directory - [AI Services](https://myrilum.cn/services): 12-domain AI service architecture - [Solutions](https://myrilum.cn/solutions): Industry and functional solution blueprints - [System Architecture](https://myrilum.cn/system): Underlying trust and verification architecture - [Regional Workbench](https://app.myrilum.cn/sign-in): Outbound authenticated MYRILUM Workbench (CN) ## AI Skills Manifold Directory - [Leaderboard](https://myrilum.cn/skills): Regional Riemannian manifold skill rankings - [Trending (24h)](https://myrilum.cn/skills/trending): 24-hour adoption velocity index - [Hot Momentum](https://myrilum.cn/skills/hot): Hourly delta and 8-week momentum - [Daily Picks](https://myrilum.cn/skills/picks): Curated daily skill highlights - [Curated Packs](https://myrilum.cn/skills/packs): Role-scoped multi-skill bundles - [Topics](https://myrilum.cn/skills/topics): Engineering and business-role taxonomies - [Supported Agents](https://myrilum.cn/skills/agents): 21 compatible IDE, CLI, and OS agents - [Model Matrix](https://myrilum.cn/skills/models): Affinity benchmarks across 6 model families - [Official Makers](https://myrilum.cn/skills/official): Verified publisher organizations - [Verified Site Sources](https://myrilum.cn/skills/sites): .well-known skill source domains - [Package Ecosystems](https://myrilum.cn/skills/ecosystems): npm, pip, cargo, and go skill packages - [Static Hygiene Audits](https://myrilum.cn/skills/audits): 3-engine security and 5-dimension hygiene checks - [Comparisons](https://myrilum.cn/skills/compare): Side-by-side paraconsistent skill trade-offs - [Documentation](https://myrilum.cn/skills/docs): CLI, Packs, API, FAQ, and customization guides ## AI Router Manifold Directory - [Router Leaderboard](https://myrilum.cn/router): Regional Riemannian manifold model routing reference baselines - [Trending Models](https://myrilum.cn/router/trending): Recency-weighted frontier benchmark momentum index (unmeasured baseline) - [Nitro Speed Tier](https://myrilum.cn/router/nitro): Highest reference decode throughput (tok/s) and lowest ref TTFT routes - [Floor Cost Tier](https://myrilum.cn/router/floor): Lowest 80/20 blended list USD/1M token pricing routes - [Inference Providers](https://myrilum.cn/router/providers): Regional and global inference provider directory - [Model Catalog](https://myrilum.cn/router/models): Full canonical model directory across domain scopes - [Model Labs](https://myrilum.cn/router/labs): Foundation model labs and origin organizations - [ZDR & Compliance Audits](https://myrilum.cn/router/audits): Fail-closed ZDR, data residency, and TLS audits - [Benchmark Suites](https://myrilum.cn/router/benchmarks): SWE-bench, LiveCodeBench, AIME, BFCL, MMMU, MRCR - [Coding Harnesses](https://myrilum.cn/router/harnesses): Compatible agent CLI, IDE, and OS runtimes - [Model Comparisons](https://myrilum.cn/router/compare): Side-by-side reference latency, value-lift, and ZDR trade-offs - [Router Documentation](https://myrilum.cn/router/docs): Unified gateway API, routing policies, BYOK, and SDKs ## Machine-Readable Indexes - [Full Skills Corpus (llms-full.txt)](https://myrilum.cn/skills/llms-full.txt) - [Skills JSON Index (v1)](https://myrilum.cn/skills/api/v1/index.json) - [Full Router Corpus (llms-full.txt)](https://myrilum.cn/router/llms-full.txt) - [Router JSON Index (v1)](https://myrilum.cn/router/api/v1/index.json) - [XML Sitemap](https://myrilum.cn/sitemap.xml) ## Top Projected Reference Models (CN) - [#1 DeepSeek V4.1 Flash](https://myrilum.cn/router/m/deepseek/deepseek-v4-1-flash): deepseek/deepseek-v4-1-flash | score=90.3 | refValue=392.08/$ | refTtft=145ms | refTps=168 | fork=CONVERGENT_LEADER - [#2 MiniMax M3.1 Flash](https://myrilum.cn/router/m/minimax/minimax-m3-1-flash): minimax/minimax-m3-1-flash | score=90.1 | refValue=381.67/$ | refTtft=138ms | refTps=172 | fork=CONVERGENT_LEADER - [#3 Zhipu GLM-5.3 Flash](https://myrilum.cn/router/m/zhipu/glm-5-3-flash): zhipu/glm-5-3-flash | score=89.9 | refValue=410/$ | refTtft=140ms | refTps=162 | fork=CONVERGENT_LEADER - [#4 MiniMax M2 Flash](https://myrilum.cn/router/m/minimax/minimax-m2-flash): minimax/minimax-m2-flash | score=86.3 | refValue=455.21/$ | refTtft=148ms | refTps=158 | fork=HIGH_VALUE_SLEEPER - [#5 MiniMax M3](https://myrilum.cn/router/m/minimax/minimax-m3): minimax/minimax-m3 | score=85.3 | refValue=195.42/$ | refTtft=215ms | refTps=124 | fork=CONVERGENT_LEADER - [#6 Qwen3 235B A22B](https://myrilum.cn/router/m/qwen/qwen3-235b-a22b): qwen/qwen3-235b-a22b | score=84.3 | refValue=317.71/$ | refTtft=178ms | refTps=136 | fork=HIGH_VALUE_SLEEPER - [#7 DeepSeek V3.2 MoE](https://myrilum.cn/router/m/deepseek/deepseek-v3-2): deepseek/deepseek-v3-2 | score=81.8 | refValue=212.84/$ | refTtft=195ms | refTps=122 | fork=CONVERGENT_LEADER - [#8 Llama 4 Maverick 400B](https://myrilum.cn/router/m/meta-llama/llama-4-maverick-400b): meta-llama/llama-4-maverick-400b | score=81.1 | refValue=262.43/$ | refTtft=162ms | refTps=164 | fork=HIGH_VALUE_SLEEPER - [#9 DeepSeek V4 Pro](https://myrilum.cn/router/m/deepseek/deepseek-v4-pro): deepseek/deepseek-v4-pro | score=80.4 | refValue=105.63/$ | refTtft=285ms | refTps=96 | fork=CONVERGENT_LEADER - [#10 MiniMax Text 01 Preview](https://myrilum.cn/router/m/minimax/minimax-text-01-preview): minimax/minimax-text-01-preview | score=79.8 | refValue=183.7/$ | refTtft=390ms | refTps=72 | fork=PRODUCTION_WORKHORSE ## Top Projected Skills (CN) - [#1 Find Skills Discovery CLI](https://myrilum.cn/skills/s/vercel-labs/skills/find-skills): vercel-labs/skills | score=87 | lift=17.83%/1k tok | fork=CONVERGENT_LEADER | best=deepseek - [#2 DeepSeek Reasoning Tree Distiller](https://myrilum.cn/skills/s/deepseek-ai/deepseek-skills/deepseek-r1-reasoning-distiller): deepseek-ai/deepseek-skills | score=84.5 | lift=19.5%/1k tok | fork=CONVERGENT_LEADER | best=deepseek - [#3 Linghe Case Brief & Scope Clarifier](https://myrilum.cn/skills/s/myrilum/linghe-skills/myrilum-case-brief): myrilum/linghe-skills | score=84 | lift=19.68%/1k tok | fork=CONVERGENT_LEADER | best=deepseek - [#4 Feishu Docs & Bitable Copilot](https://myrilum.cn/skills/s/feishu-open/feishu-agent-skills/feishu-doc-copilot): feishu-open/feishu-agent-skills | score=82.5 | lift=15%/1k tok | fork=PRODUCTION_WORKHORSE | best=deepseek - [#5 GEO Citation & llms.txt Auditor](https://myrilum.cn/skills/s/myrilum/linghe-skills/myrilum-geo-citation-auditor): myrilum/linghe-skills | score=81.5 | lift=18.47%/1k tok | fork=HIGH_LIFT_SLEEPER | best=deepseek - [#6 Commercial Contract Redline Sentinel](https://myrilum.cn/skills/s/myrilum/enterprise-skills/myrilum-contract-redline-sentinel): myrilum/enterprise-skills | score=81.2 | lift=26.31%/1k tok | fork=HIGH_LIFT_SLEEPER | best=deepseek - [#7 MCP Slim Schema Validator](https://myrilum.cn/skills/s/modelcontextprotocol/registry-skills/mcp-schema-validation-slim): modelcontextprotocol/registry-skills | score=81.2 | lift=27.31%/1k tok | fork=HIGH_LIFT_SLEEPER | best=deepseek - [#8 Context Token Efficiency Probe](https://myrilum.cn/skills/s/vercel-labs/skills/skill-telemetry-probe): vercel-labs/skills | score=80.7 | lift=27.96%/1k tok | fork=HIGH_LIFT_SLEEPER | best=deepseek - [#9 Vercel React & RSC Best Practices](https://myrilum.cn/skills/s/vercel-labs/agent-skills/vercel-react-best-practices): vercel-labs/agent-skills | score=80.2 | lift=15.41%/1k tok | fork=CONVERGENT_LEADER | best=qwen - [#10 CN Static Export & ICP Preflight](https://myrilum.cn/skills/s/aliyun-bailian/bailian-skills/cn-icp-filing-preflight): aliyun-bailian/bailian-skills | score=80.2 | lift=27.06%/1k tok | fork=HIGH_LIFT_SLEEPER | best=deepseek - [#11 Qwen React SSR & Mobile Viewport Kit](https://myrilum.cn/skills/s/qwen-lm/qwen-agent-skills/qwen-react-ssr-kit): qwen-lm/qwen-agent-skills | score=79.9 | lift=14.25%/1k tok | fork=PRODUCTION_WORKHORSE | best=qwen - [#12 QQ Agent Bot Router & Handover](https://myrilum.cn/skills/s/tencent-agent/qq-agent-skills/qq-enterprise-bot-router): tencent-agent/qq-agent-skills | score=79.2 | lift=12.97%/1k tok | fork=PRODUCTION_WORKHORSE | best=deepseek ## Topic Collections - [React & UI Architecture (React 与组件架构)](https://myrilum.cn/skills/topics/react): React 19 concurrent rendering, component boundaries, and hydration controls. - [Next.js App Router (Next.js 全栈路由)](https://myrilum.cn/skills/topics/nextjs): App Router static params, canonical URL alignment, and static export safety. - [Design Systems & Tokens (设计系统与界面令牌)](https://myrilum.cn/skills/topics/design): Token-governed design systems, contrast parity, and layout discipline. - [Mobile & Mini-App UX (移动端与多端适配)](https://myrilum.cn/skills/topics/mobile): Mobile reflow, mini-app payloads, and zero horizontal overflow rules. - [Agent Workflows & MCP (智能体工作流与 MCP)](https://myrilum.cn/skills/topics/agent-workflows): MCP tool schemas, skills lockfiles, sandbox gates, and eval harnesses. - [Databases, RLS & Vector (数据库、RLS 与向量检索)](https://myrilum.cn/skills/topics/databases): Postgres RLS policies, edge SQL, hybrid retrieval, and safe DDL migrations. - [Testing & Verification Gates (自动化测试与验证门禁)](https://myrilum.cn/skills/topics/testing): Deterministic E2E suites, tool-call evals, and reproducible CI runners. - [Growth Engineering & GEO (增长工程与 GEO 索引)](https://myrilum.cn/skills/topics/marketing): llms.txt curation, Schema.org Dataset JSON-LD, and bilingual copy checks. - [Strategic Research & Scoping (战略研究与需求澄清)](https://myrilum.cn/skills/topics/strategic-research): Goal diagnosis, case briefs, reasoning trees, and enterprise RAG sync. - [Commercial & Legal Review (商业交易与风控法务)](https://myrilum.cn/skills/topics/commercial-legal): Contract clause diffing, RFP matrices, and ledger reconciliation review. - [Marketing & Media Delivery (内容增长与多模态传播)](https://myrilum.cn/skills/topics/marketing-media): GEO citation audits, editorial checks, and enterprise social routing. - [Compliance & Cell Governance (合规备案与双区治理)](https://myrilum.cn/skills/topics/compliance-filing): PIPL/GDPR data-flow checks, CN static export preflight, and SPDX audits. - [Digital Engineering Delivery (数字工程全链交付)](https://myrilum.cn/skills/topics/digital-engineering): Core full-stack architecture, security guards, and design-system gates. ## Model Families - [DeepSeek](https://myrilum.cn/skills/models/deepseek) (DeepSeek) - [Qwen](https://myrilum.cn/skills/models/qwen) (Alibaba Cloud) - [Doubao](https://myrilum.cn/skills/models/doubao) (ByteDance) - [Claude](https://myrilum.cn/skills/models/claude) (Anthropic) - [Gemini](https://myrilum.cn/skills/models/gemini) (Google) - [GPT](https://myrilum.cn/skills/models/gpt) (OpenAI)