Analyze a codebase to produce an interactive knowledge graph for understanding architecture, components, and relationships
Skills(SKILL.md)は、AIエージェント(Claude Code、Cursor、Codexなど)に特定の能力を追加するための設定ファイルです。
詳しく見る →Analyze a codebase to produce an interactive knowledge graph for understanding architecture, components, and relationships
Designs and implements production-grade RAG systems by chunking documents, generating embeddings, configuring vector stores, building hybrid search pipelines, applying reranking, and evaluating retrieval quality. Use when building RAG systems, vector databases, or knowledge-grounded AI applications requiring semantic search, document retrieval, context augmentation, similarity search, or embedding-based indexing.
State-space model with O(n) complexity vs Transformers' O(n²). 5× faster inference, million-token sequences, no KV cache. Selective SSM with hardware-aware design. Mamba-1 (d_state=16) and Mamba-2 (d_state=128, multi-head). Models 130M-2.8B on HuggingFace.
RNN+Transformer hybrid with O(n) inference. Linear time, infinite context, no KV cache. Train like GPT (parallel), infer like RNN (sequential). Linux Foundation AI project. Production at Windows, Office, NeMo. RWKV-7 (March 2025). Models up to 14B parameters.
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.
Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.
Serverless GPU cloud platform for running ML workloads. Use when you need on-demand GPU access without infrastructure management, deploying ML models as APIs, or running batch jobs with automatic scaling.
Provides guidance for experiment tracking with SwanLab. Use when you need open-source run tracking, local or self-hosted dashboards, and lightweight media logging for ML workflows.
Agentic Workflow Pattern
Guide to the math cognitive stack - what tools exist and when to use each
Deterministic router for math cognitive stack - maps user intent to exact CLI commands
Token-efficient code analysis via 5-layer stack (AST, Call Graph, CFG, DFG, PDG). 95% token savings.
Create a customer journey map across stages, touchpoints, actions, emotions, and metrics. Use when diagnosing a broken experience or aligning a team on the full customer flow.
Break down epics into user stories with Humanizing Work split patterns. Use when a backlog item is too large to estimate, sequence, or deliver safely.
Create a user story map that lays out activities, steps, tasks, and release slices. Use when planning a workflow, backlog, or MVP around the user journey.
Zero-config deployment tool: upload static files to IPFS, or create and deploy full-stack web projects (React+Vite + Cloudflare Worker + D1 database). Workers also support sending emails via the PinMe
Sonnet Amplified fullstack engine. 34 modes, SEC-01~15 OWASP security, 13 runtime hooks, 75% token reduction. Install: npx @smorky85/aurakit
Automate Box cloud storage operations including file upload/download, search, folder management, sharing, collaborations, and metadata queries via Rube MCP (Composio). Always search tools first for current schemas.
Deploy a free VLESS proxy/VPN node on Cloudflare Pages using edgetunnel. Automates code download, UUID generation, Pages deployment, free domain registration (DNSExit), DNS configuration, custom domain binding, and client setup for Shadowrocket/v2rayN/Clash. Uses Cloudflare Pages (not Workers) because Pages supports CNAME-based custom domains from any DNS provider, avoiding the need to host DNS on Cloudflare.
Automate Datadog tasks via Rube MCP (Composio): query metrics, search logs, manage monitors/dashboards, create events and downtimes. Always search tools first for current schemas.
Automate Dropbox file management, sharing, search, uploads, downloads, and folder operations via Rube MCP (Composio). Always search tools first for current schemas.
Intelligently organizes your files and folders across your computer by understanding context, finding duplicates, suggesting better structures, and automating cleanup tasks. Reduces cognitive load and keeps your digital workspace tidy without manual effort.
Automate GitLab project management, issues, merge requests, pipelines, branches, and user operations via Rube MCP (Composio). Always search tools first for current schemas.
Automate Google Drive file operations (upload, download, search, share, organize) via Rube MCP (Composio). Upload/download files, manage folders, share with permissions, and search across drives programmatically.
>
Automate HubSpot CRM operations (contacts, companies, deals, tickets, properties) via Rube MCP using Composio integration.
Automate OneDrive file management, search, uploads, downloads, sharing, permissions, and folder operations via Rube MCP (Composio). Always search tools first for current schemas.
Automate Pipedrive CRM operations including deals, contacts, organizations, activities, notes, and pipeline management via Rube MCP (Composio). Always search tools first for current schemas.
AI-powered browser automation — navigate sites, fill forms, extract structured data, log in with stored credentials, and build reusable multi-step workflows using natural language. Install: pip install skyvern && skyvern setup
Deploy status across all projects. Shows ECS service versions, Vercel deployments, recent deploys, pending deploys, and CI/CD pipeline state.
Health check and auto-repair for the ops plugin. Diagnoses manifest errors, broken permissions, invalid configs, stale caches, and missing files — then spawns an agent to fix everything automatically.
Shopify store command center. Orders, inventory, fulfillment, analytics, and store health. Works with any Shopify store via Admin API.
Production incidents dashboard. Reads ECS health, Sentry errors, CI failures. Offers to dispatch fix agents for active fires.
Marketing command center. Email campaigns (Klaviyo), paid ads (Meta/Google), analytics (GA4), SEO, and social media metrics. One dashboard for all marketing channels.
Before advising, load:
Autonomous multi-project orchestration engine. Audits all registered projects, structures work into dependency-wired tasks, dispatches parallel agents (subagents or Agent Teams), audits completions, and ships PRs. Registry-driven — works for any user with a configured project registry.
Business operations command center. Routes to the right ops command based on what you need — briefing, inbox, fires, projects, comms, triage, linear, revenue, deploy, or yolo mode.
Create visual moodboards from collected inspiration with iterative refinement. Use after trend research or website analysis to synthesize design direction before implementation.
Browse and select fonts from Google Fonts or curated pairings. Use to find the perfect typography for a design project.
Add phase to end of current milestone in roadmap
Capture idea or task as todo from current conversation context
Analyze phase dependencies and suggest Depends on entries for ROADMAP.md
List pending todos and select one to work on
Gather phase context through adaptive questioning before planning. Use --auto to skip interactive questions (Claude picks recommended defaults). Use --chain for interactive discuss followed by automatic plan+execute. Use --power for bulk question generation into a file-based UI (answer at your own pace).
Execute all plans in a phase with wave-based parallelization
Interactive command center for managing multiple phases from one terminal
Resume work from previous session with full context restoration
Display project statistics — phases, plans, requirements, git metrics, and timeline
You are the master orchestrator for the vibe-coding workflow. Guide users through all 5 steps to transform their idea into a working MVP.
Configure file watching hooks to auto-react to config changes, env file updates, and dependency modifications. Use to set up reactive workflows.