发现 Skills

Software surgery protocol for proof-backed, minimal-diff implementation: Contract → Invariants → Creative Frame → inevitable Incision → Proof. Use when users say "run $tk", "patch-first", "keep the diff small", "state contract + invariants", "one clear incision", "fix it with a validation signal", or orchestration cues like "workers use $tk" / "internally use $tk".

Applies Anthropic's official brand colors and typography to any artifact that may benefit from Anthropic's look and feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.

Add phase to end of current milestone in roadmap

A highly specialized skill designed to audit and optimize a Territorial Intelligence NotebookLM workspace (specifically focused on Southwest Mallorca Real Estate, Luxury Market, Tourism, and Urban Development).

Specialized skill to audit, optimize, and maintain the NotebookLM workspace for "OpenClaw: The Universal Personal AI Assistant Platform". Use this skill when you need to: (1) audit or clean sources in the main OpenClaw notebook or any of its 4 sub-notebooks, (2) remove duplicate, obsolete, or low-value sources from any OpenClaw NotebookLM notebook, (3) configure or update the AI persona/instructions for any OpenClaw sub-notebook, (4) validate the retrieval quality of OpenClaw knowledge bases, (5) do any deep research, pruning, or restructuring task related to the OpenClaw NotebookLM workspace.

Enforce KISS, Purity, SOLID, and Let It Crash principles through mandatory validation gates. Detects complexity violations, impure functions, design anti-patterns, and error handling issues.

Use when user requests Home Assistant YAML automations, blueprints, scripts, or config. MANDATORY: First response must be clarifying questions, NEVER code. NOT for: Node-RED flows (use node-red skill), device firmware (use esphome skill).

Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.

Property valuation, market analysis, investment ROI calculations, comparable analysis, and rental yield assessment. Use when evaluating real estate investments, analyzing property markets, or calculating returns.

个人认知框架系统。当涉及目标规划、决策建议、项目方向、工作方式优化时触发，或用户显式调用"/telos"。加载完整画像提供个性化响应，支持画像进化和定期复盘。

Use when building AI agents with Microsoft.SemanticKernel or Microsoft.Extensions.AI.Agent frameworks. Covers agent creation, tool/plugin registration, multi-agent orchestration, chat completion agents, and OpenAI Assistants integration. USE FOR: building AI agents with tools and plugins, multi-agent chat workflows, Semantic Kernel agent orchestration, OpenAI Assistants API integration, agent-based automation pipelines DO NOT USE FOR: simple chat completions without agent patterns (use microsoft-extensions-ai), inter-service agent communication over A2A protocol (use a2a), ML model training (use mlnet), raw ONNX inference (use onnx)

Identical Twins Protocol — compressed agent-to-agent messaging. Reduces inter-agent token usage 40-80% via shared codebook. Use when sending messages between Claude instances.

Guides agents through analytics engineering workflows with dbt. Use when building or modifying staging models, marts, tests, snapshots, documentation, exposures, or semantic-layer-facing models.

Operational patterns for LLM inference: latency budgeting, tail-latency control, caching, batching/scheduling, quantization/compression, parallelism, and reliable serving at scale. Emphasizes production-grade performance, cost control, and observability.

Use when building custom AEO monitoring infrastructure. Covers API access, scraping architecture, legal compliance, and cost estimation for AI search visibility tracking. Reference implementation: TypeScript/Next.js/Supabase stack.

Production-grade backend service development across Node.js (Express/Fastify/NestJS/Hono), Bun, Python (FastAPI), Go, and Rust (Axum), with PostgreSQL and common ORMs (Prisma/Drizzle/SQLAlchemy/GORM/SeaORM). Use for REST/GraphQL/tRPC APIs, auth (OIDC/OAuth), caching, background jobs, observability (OpenTelemetry), testing, deployment readiness, and zero-trust defaults.

Use when working on Slack agent/bot code, Chat SDK applications, Bolt for JavaScript projects, or projects using @chat-adapter/slack or @slack/bolt. Provides development patterns, testing requirements, and quality standards.

name: ncc description: Expert guidance for @vercel/ncc — a simple CLI for compiling Node.js modules into a single file with all dependencies included. Use when bundling serverless functions, CLI to...

Expert guidance for Geist, Vercel's default typography system and font family for precise Next.js interfaces. Use when configuring Geist Sans, Geist Mono, or Geist Pixel, setting up font imports, or applying Vercel typography and aesthetic guidance.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.

Use when writing, reviewing, or refactoring Terraform HCL — covers project setup, resource authoring, module development, and pre-commit validation.

Multi-phase workflow orchestration pattern for main agents. Enables structured planning, delegation to sub-agents, parallel execution, and result aggregation.

Use this skill when managing VMOS Edge cloud phone containers through the Container API via the host machine IP (`host_ip`). Covers host queries, instance lifecycle, db_id/cloud_ip lookup, batch operations, app distribution, and async polling. If no IP is provided and local `cbs_go` is running, default `host_ip` to `127.0.0.1`. Use it to resolve `db_id` when `vmos-edge-control-api` needs host_ip-based routing.