Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Development
Use this skill when deciding whether font-size should use rem or px, reviewing Chrome text zoom support, or choosing whether text-dependent spacing and sizing should scale with typography. Apply it when CSS uses blanket rem conversion, root font-size hacks, or relative units such as em and ch without a clear semantic basis.
Run any pending knowledge-base migration by querying the deterministic migration chain (`migrate status`) and executing each pending step's documented procedure in-host, with every write delegated to the step's deterministic CLI primitives. Use when the node reader, `doctor`, or `init` reports an out-of-date `schema_version` / legacy flat layout and asks you to migrate, or when the user asks to migrate the knowledge base.
Bootstrap or extend Linear project tracking the right way — methodology, not API CRUD. Use when the user asks to set up Linear tracking, track this project in Linear, add this repo to Linear, backfill history, get ready to track, or consents after a soft-nudge from linear-discipline. Anchors to North Star, backfills Done with evidence, wires blockedBy, installs per-repo discipline. NOT for one-off issue create/comment — Linear MCP handles that.
Animation, morphing, and shader best practices for Liquid Glass and SwiftUI motion on iOS 26+/macOS 26+ through iOS 27. Use when animating glass or building UI motion: glassEffectID, @Namespace morphing, GlassEffectContainer spacing, glassEffectUnion, glassEffectTransition (.matchedGeometry/.materialize), withAnimation, springs, UnitCurve, PhaseAnimator, KeyframeAnimator, @Animatable, matchedTransitionSource, navigationTransition zoom, scrollTransition, visualEffect, Metal shaders (colorEffect, distortionEffect, layerEffect, [[stitchable]]), accessibilityReduceMotion gating, Instruments/xctrace profiling of hangs and animation hitches. Also fires on symptoms: glass morph not animating, view pops in instead of morphing, glass fades instead of materializing, morph flickers with Reduce Motion, menu glass morphs as a rectangle, animation runs on insert but not removal, PhaseAnimator loops forever or fires once, shader shows clipped edges (maxSampleOffset), animation hitches or main-thread hangs while scrolling glass chrome. For static glass, availability gating, toolbars, UIKit/AppKit glass, or accessibility of the material itself use the liquid-glass skill first.
Implement large language model (LLM) chat completions using the z-ai-web-dev-sdk. Use this skill when the user needs to build conversational AI applications, chatbots, AI assistants, or any text generation features. Supports multi-turn conversations, system prompts, and context management.
Use when integrating the Ludeo SDK into an Unreal Engine game, performing Ludeo integration analysis, designing or implementing Ludeo integration architecture in UE code, or working on any phase of a Ludeo SDK integration (lifecycle, actions, state tracking, player flow). Also trigger when the user mentions Ludeo SDK, Ludeo integration, playable highlights integration, LudeoSession, LudeoRoom, DataWriter, DataReader, or game state tracking for Ludeo - when the work is on an Unreal Engine game's code. Do NOT use for Unity or non-Unreal engines (use ludeo-unity-integration instead), pure SDK-concept/documentation questions with no code work, cloud cast session/VM log diagnosis (use ludeo-diagnose-session), repo open-sourcing, or creating decks/docs about an integration.
Control Luvus through its local CLI and UHP. Use only for a line beginning with `=target message`, an explicit request naming Luvus, a request to delegate to a named live Luvus agent or pane, or an explicit Luvus operation involving sessions, machines, workspaces, tabs, panes, agents, files, Git, DIFF, worktrees, tasks, leases, modules, themes, Luvus Bar, configuration, UI, integrations, or Luvus UHP. Do not use for ordinary coding, file edits, Git operations, tests, task planning, generic agent work, or parallelization unless the user explicitly connects the request to Luvus. Being inside Luvus does not trigger this skill by itself. Inside Luvus use the inherited session; outside use the installed production Luvus command and configured session.
Git CLI operations, workflows, and automation for modern development (2025). Use when working with repositories, commits, branches, merging, rebasing, worktrees, submodules, or multi-repo architectures. Includes parallel agent workflow patterns, merge strategies, conflict resolution, and large repo optimization. Triggers on git commands, version control, merge conflicts, worktree setup, submodule management, repository troubleshooting, branch strategy, rebase operations, cherry-pick decisions, and CI/CD git integration.
Provides comprehensive guidance for MateChat Vue 3 component library including AI chat/agent UI components, installation, API reference, and demo examples. Use when building AI chat interfaces with Vue 3, implementing conversation bubbles, input components, or any MateChat UI components.
Optimize MATLAB design files for GPU Coder to generate faster CUDA code. Iteratively profiles, rewrites, and benchmarks until performance targets are met or diagnostics are resolved. Use when asked to: optimize for GPU Coder, improve GPU codegen performance, profile generated GPU/CUDA code, profile GPU MEX, fix gpuPerformanceAnalyzer diagnostics, speed up GPU MEX, reduce GPU memory transfers, improve kernel parallelism, rewrite MATLAB for CUDA, or run gpuPerformanceAnalyzer.
Store, retrieve, and pass credentials securely in MATLAB using the built-in MATLAB Vault (setSecret, getSecret, importSecrets, secretID) instead of hardcoding. Handles connections to any authenticated service: REST APIs, databases, cloud storage (S3/Azure/GCS), SFTP, and others. Covers API keys, tokens, passwords, SSH passphrases, CI/batch/scheduled jobs, and "keep credentials out of code" requests. Does NOT cover third-party secret managers (HashiCorp Vault, AWS Secrets Manager), OS-level key management, or the connection/query logic itself.
How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.
Validate or decompose any specification, plan, or requirement list using the MECE framework (Mutually Exclusive, Collectively Exhaustive). Use when reviewing PRDs, feature lists, user stories, API designs, state management, roadmaps, or any structured breakdown. Also use when the user says "check this", "anything missing?", "is this complete?", or "break this down".
Guides Metronome usage-based billing integration decisions — event ingestion (single and batch, idempotency, billable metrics), contract design (rate cards, overrides, dimensional pricing, products), invoicing lifecycle (grace periods, finalization, Stripe sync), credit and commit management (prepaid, postpaid, thresholds, auto-recharge), and Stripe integration (arrears invoicing, tax providers, line item limits). Use when building, modifying, or reviewing any Metronome integration — including ingesting usage events, creating contracts or rate cards, managing credits and commits, configuring invoicing, or syncing invoices with Stripe Billing.
Use when the user wants an autonomous ML research loop that does more than blindly try changes. After every training run the agent analyses what actually happened inside the model — gradients, activations, embeddings, errors, data — and grounds the next change in that evidence. A `<literature>` on/off dial adds scientific-literature grounding: off behaves as a pure analysis-first loop; on searches papers, grades the evidence, and implements only what prior work supports. One change per run; loops forever until interrupted. Not for one-off training runs or hyperparameter sweeps.
Design and audit AI model routers that balance task quality, latency, cost, reliability, safety, privacy, context size, tool use, fallback, caching, evals, budgets, and provider outages. Use when an agent or product must choose models dynamically instead of hard-coding one default model.
Guide for modernizing legacy Python 2 scientific computing code to Python 3 with modern libraries. This skill should be used when migrating scientific scripts involving data processing, numerical computation, or analysis from Python 2 to Python 3, or when updating deprecated scientific computing patterns to modern equivalents (pandas, numpy, pathlib).
Automatically activates when over-engineering patterns are detected: (1) Modifying code or files the user did not explicitly ask to change (2) Creating new abstraction layers (class, interface, factory, wrapper) without being asked (3) Adding comments, documentation, JSDoc, or type annotations without being asked (4) Introducing new dependencies without being asked (5) Rewriting entire files instead of making minimal edits (6) Diff scope significantly exceeding the user's request (7) User signals like "too much", "don't change that", "only change X", "keep it simple", "stop" (8) Adding error handling, validation, or defensive code for scenarios that cannot occur (9) Generating tests, configuration scaffolding, or documentation without being asked
Convert single-node scripts to multi-node Slurm sbatch jobs and debug common multi-node failures. Covers srun-native vs uv run torch.distributed approaches, container setup, NCCL timeouts, OOM sizing for MoE models, and interactive allocation
Systematic workflow for MoE training optimization in Megatron Bridge, based on the Megatron-Core MoE paper. Covers the Three Walls framework, parallel folding, recompute strategy, dispatcher choice, and CUDA-graph bring-up.
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.