Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Cloud & DevOps
Diagnose and prevent `vbar_control_agent` segfaults and OOMs caused by race conditions during TPU device resets and frequent metrics collection (e.g. every 3s). Use when TPU slice initialization fails or `vbar_control_agent` crashes on TPU v6e nodes.
Guide for creating GKE ComputeClass resources. Use this skill when users want to define custom node configurations, autoscaling priorities, or hardware requirements (e.g., Spot VMs, GPUs, specific machine families) for their GKE workloads.
Manage GoCD pipelines, pipeline groups, agents, environments, config repos, server administration, users, roles, authorization configs, plugins, backups, materials, artifact stores, elastic agent profiles, cluster profiles, stages, jobs, server configuration, templates, packages, package repositories, notification filters, dashboard, access tokens, secret configs, and server version using the orbit CLI. Use this skill whenever the user asks about GoCD pipelines, agents, environments, config repos, server health, maintenance mode, CI/CD operations, pipeline groups, users, roles, auth configs, plugins, backups, materials, artifact stores, elastic agents, stages, jobs, templates, packages, package repositories, notification filters, dashboard, access tokens, secret configs, or server version on GoCD. Trigger on phrases like 'list pipelines', 'create pipeline', 'delete pipeline', 'pipeline status', 'pipeline config', 'pipeline materials', 'what branch', 'trigger a build', 'pause pipeline', 'list agents', 'enable agent', 'disable agent', 'kill running tasks', 'agent job history', 'list environments', 'create environment', 'patch environment', 'config repo status', 'create config repo', 'preflight check', 'server health', 'maintenance mode', 'encrypt a value', 'list users', 'create user', 'list roles', 'auth config', 'list plugins', 'schedule backup', 'list materials', 'artifact store', 'elastic agent profile', 'cluster profile', 'cancel stage', 'run stage', 'run job', 'site url', 'job timeout', 'mail server config', 'pipeline group', 'list templates', 'create template', 'dashboard', 'access tokens', 'secret config', 'compare pipelines', 'lock pipeline', 'unlock pipeline', 'server version', 'notification filters', 'list packages', 'package repository', 'current user', or any GoCD-related task — even casual references like 'what pipelines are running', 'is the agent idle', 'check the build', 'schedule a run', 'put server in maintenance', 'check config repo sync', 'who has access', 'what plugins are installed', or 'GoCD status'. The orbit CLI alias is `cd`.
Deploy a self-written app as a Home Assistant OS (HAOS) local add-on (now labelled 'app' in the HA UI/CLI) on a Raspberry Pi for 24/7 operation, e.g. moving a poller or monitor off a workstation that sleeps. Use when deploying, updating or debugging a HAOS local add-on / HA app, or when the user mentions 部署到 RPi / HAOS / HA add-on / HA app. Also consult when an add-on update is not detected (update_available stays false, version_latest shows the old version) or a rebuild keeps running old code.
Investigate security incidents in Microsoft Azure (resource and subscription control plane) — reconstruct attacker activity from the Azure Activity Log and resource/data-plane diagnostic logs, anchor the investigation on the identity that made the calls (a user, service principal, or managed identity), trace privilege escalation through role assignments, hunt managed-identity token abuse and VM run-command code execution, and detect storage or Key Vault data theft while correlating back to Entra sign-in logs. Use when responding to a suspected Azure resource compromise, anomalous Azure Activity Log entries, a Microsoft Defender for Cloud alert, managed-identity or service-principal abuse, a crypto-mining VM, or storage-account exfiltration.
Investigate a security incident in Google Cloud — establishing what audit logging exists before trusting a gap, reconstructing activity from Cloud Audit Logs, triaging service-account and OAuth abuse, following Security Command Center findings, and scoping IAM and resource changes. Use when responding to a suspected GCP compromise, investigating a leaked service-account key, working a Security Command Center or Event Threat Detection alert, or reconstructing what a principal did across a GCP organization
Plans LinuxMirrors Docker Engine installation, Docker CE package source changes and Docker Registry Mirror configuration for Linux hosts, covering version selection, existing installations, registry-only changes, multiple mirror addresses, daemon.json merging, service verification and rollback. Use when the user mentions docker.sh, DockerInstallation.sh, Docker pull acceleration or `--only-registry`; do not use this skill for ordinary system package source changes.
Create or modify GitHub Actions workflows for CI, automation, or PR management. Use when asked to create, scaffold, add, extend, or fix a workflow (.yml file under .github/workflows/), including existing pull-request policy and comment automation.
Use when managing BK-CI pipeline builds via MCP, such as querying build history, getting launch parameters, viewing build status, and triggering builds after confirmation. Prefer this when the user wants to operate existing pipelines rather than modify code.
Guided workflow for taking any ML codebase — including one with no experiment tracking at all, or one full of MLflow 2-era idioms — to a production-grade open-source MLflow 3 setup with dev/staging/prod environments, registry-based promotion, and served models. Walks seven phases with a developer who may have zero MLflow 3 experience — assess the codebase (scripted read-only audit), model the registry domain (per-environment model names, aliases, gates), stand up tracking per environment, restructure training code to MLflow 3 idioms, wire evaluation-gated promotion, serve and smoke-test, then run the ongoing MLOps loop. Use when asked to set up MLflow, migrate to MLflow 3, productionize model training and serving, or design a dev/staging/prod MLOps cycle. Pairs with the sibling mlflow-3 rule pack for every API decision.
Master monorepo management with Turborepo, Nx, and pnpm workspaces to build efficient, scalable multi-package repositories with optimized builds and dependency management. Use when setting up monorepos, optimizing builds, or managing shared dependencies
Set up serverless Postgres with Neon or Vercel Postgres for Cloudflare Workers/Edge. Includes connection pooling, git-like branching for preview environments, and Drizzle/Prisma integration. Use when: setting up edge Postgres, configuring database branching, or troubleshooting "TCP not supported", connection pool exhausted, SSL config (sslmode=require), or Prisma edge compatibility.
Improve Coval trace quality after basic ingestion works. Use when traces are sparse, missing useful STT/LLM/TTS/tool spans, missing attributes needed for Coval built-in metrics, or when a customer wants maximum debugging and observability value from agent traces.
Use when operating Ultralytics Platform from a terminal or script with the ul cloud CLI (ul cloud RESOURCE OPERATION key=value, PyPI package ultralytics-platform) — listing, creating, updating, cloning, or moving projects, datasets, and models; inspecting dataset images, statistics, metadata, duplicates, and similar images; starting and monitoring cloud training runs; per-image model validation analysis; downloading weights; exports, deployments, uploads, trash and restore, storage integrations, and account usage, storage, and billing; and building Agents (YOLO, LLM, gate, and alert workflows). For the Platform web UI or local yolo commands, see yolo; for choosing what to train and how to improve it, see yolo-training and yolo-tuning.
LLM workloads on open-source Ray (pinned to 2.57) — OpenAI-compatible serving with ray.serve.llm (vLLM-backed LLMConfig + build_openai_app) and batch inference with ray.data.llm (build_processor). Corrects the stale defaults a model produces — the archived ray-llm repo and its YAML configs, hand-rolled vLLM engines inside plain Serve deployments, the removed build_llm_processor name, deprecated boolean stage flags, top-level LLMServer/LLMRouter imports, free-form accelerator strings, one-deployment-per-LoRA-adapter designs — with the 2.57 idioms (stage configs, placement_group_config over hand-rolled PGs, deployment_config autoscaling, dynamic LoRA multiplexing, prefix-cache-affinity routing, the full OpenAI endpoint surface). Use when writing, reviewing, or productionizing LLM serving or batch inference on Ray. Classic-ML Ray (Train/Tune/Data/Serve/clusters) lives in the sibling ray skill.
Design and audit release health dashboards, rollout gates, canary metrics, crash/error monitoring, support signals, store release status, desktop auto-update adoption, rollback triggers, and post-release review for apps, SaaS, games, web, and desktop software. Use when launching releases, patches, experiments, or staged rollouts that need operational confidence.
Runs an RHDH nightly ProwJob on demand through the OpenShift CI Gangway REST API, for `periodic-ci-redhat-developer-rhdh-*-nightly` and `periodic-ci-redhat-developer-rhdh-plugin-export-overlays-*-nightly` jobs, with optional image registry/repo/tag, catalog-index, Helm chart, Playwright, fork, and Slack-alert overrides. Use for "trigger the nightly", "run the e2e job now", "kick off the AKS operator job on 1.9", RC or GA image verification runs, or listing the available nightly jobs and quay.io tags.
Use for Roblox Open Cloud and external integration work: choosing API-key-based authentication for server-to-server automation, constructing and troubleshooting Open Cloud REST requests, checking scopes and rate limits, determining whether HttpService can call an endpoint, setting up webhook receivers, and using Roblox OpenAPI and reference JSON artifacts for tooling or client generation.
Set up Sentry releases and deploy tracking — tag events with a version and environment, create the release in CI with its commits, and wire up suspect commits and code mappings, so Sentry can show which release introduced an issue, which commit is responsible, and release health. Use when asked to set up releases, track deploys, see what changed, or when issues show an unknown release or no suspect commit.
Build full-stack React applications with TanStack Start on Cloudflare Workers. Type-safe routing, server functions, SSR/streaming, and seamless D1/KV/R2 integration. Use when: building full-stack React apps, need SSR with Cloudflare Workers, want type-safe server functions, or migrating from Next.js. RC status - v1.0 stable pending. Monitor tanstack/router#5734 (memory leak) before production use.
Workflows for generating terraform solution that are the composition of one or several Terraform IBM Modules (TIM). Use when working with IBM Cloud infrastructure as code, Terraform modules, infrastructure automation, or cloud resource provisioning. Provides workflows for module discovery, composition patterns, code generation, and validation. Essential for tasks involving IBM Cloud VPC, compute, networking, security, databases, observability, or any IBM Cloud service deployment. Triggers on keywords like "terraform", "IBM Cloud", "infrastructure", "IaC", "modules", "deploy", "provision", or specific IBM Cloud services (VPC, VSI, OpenShift, etc.).
Integrate Redis-compatible Vercel KV for caching, session management, and rate limiting in Next.js applications. Powered by Upstash with strong consistency and TTL support. Use when implementing cache strategies, storing temporary data with expiration, building rate limiters, or troubleshooting missing environment variables, serialization errors, or rate limit issues.
Install and configure vLLM-Plugin-FL for multiple hardware backends including NVIDIA, Ascend and etc. Use when setting up vllm-plugin-fl, configuring the environment for specific hardware backend, installing dependencies, checking whether dependencies are installed successfully, resolving runtime issues, and launching inference to verify successful model serving. Trigger when the user says things like "setup vllm-plugin-fl", "install vllm-plugin-fl", "configure FL plugin", "set up FlagGems", or "set up FlagCX".
Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.
The skill should be used when the user asks about Alibaba Cloud AgentLoop platform for onboarding applications into observability, high-code instrumentation with loongsuite-genai-utils and OpenTelemetry SDK (高代码埋点、LLM Trace 字段、上下文传递与链路串联), Live-Debug runtime diagnostics, managing Datasets, building pipelines, and evaluating. Live-Debug covers ServiceTask dynamic logging, snapshots, metrics, spans, and JVM inspection.
Read-only diagnostics for Alibaba Cloud CDN refresh and preload issues. Use when a URL/file/directory refresh or preload looks ineffective - refresh failed, preload failed, cache not cleared, or the task failed. Verifies task records and edge cache status, and produces a diagnosis report; never submits refresh/preload jobs. Triggers: "refresh failed", "preload failed", "cache not cleared", "purge not working", "prefetch not cached", "warm up ineffective", "invalidation unsuccessful", "pre-warming failure", "pre-fetching".
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.