Skills
98,873 packaged capabilities and playbooks for agents. Install any of them with one command.
All
Use this skill for Node.js backend AI via @cloudbase/node-sdk (>=3.16.0) — cloud functions, CloudRun, Express/Koa/NestJS, serverless APIs, scheduled jobs, LLM proxies, agent orchestration. The only SDK supporting image generation (ai.createImageModel + generateImage). Text via ai.createModel with groups cloudbase, the legacy hunyuan-v3 / hunyuan-exp, or custom-*; model ids (e.g. deepseek-v4-flash, glm-5, kimi-k2.6) go in the `model` field of generateText/streamText. MUST run two-step preflight before code — see body. NOT for browser/Web (use ai-model-web) or Mini Program (use ai-model-wechat).
Open-source embedding database for AI applications. Store embeddings and metadata, perform vector and full-text search, filter by metadata. Simple 4-function API. Scales from notebooks to production clusters. Use for semantic search, RAG applications, or document retrieval. Best for local development and open-source projects.
GDB debugger skill for C/C++ programs. Use when starting a GDB session, setting breakpoints, stepping through code, inspecting variables, debugging crashes, using reverse debugging (record/replay), remote debugging with gdbserver, or loading core dumps. Activates on queries about GDB commands, segfaults, hangs, watchpoints, conditional breakpoints, pretty-printers, Python GDB scripting, or multi-threaded debugging.
An orchestrator skill that polishes Korean text written by AI (ChatGPT, Claude, Gemini, etc.) so it reads "as if a person wrote it". It detects and classifies 85 AI patterns across 10 categories — translationese, excessive English quotations, mechanical parallelism, clichés, overuse of passive voice, overuse of connectives, uniform rhythm, too many emoji/bullets, and more — and rewrites only style, rhythm and expression, leaving the content untouched. The shim's route_hint(light|standard|heavy) decides the path: well-written text takes 1 call, standard 2 calls, and only severe or long text needs 3+ calls (diagnose → targeted polish → finalize). Triggers — "remove the AI feel", "make AI text sound natural", "GPT/ChatGPT style", "fix AI translationese", "polish it like a human wrote it", "AI polish", "remove ChatGPT feel", "detect and polish AI text in Korean", "make AI text read like a human", "remove translationese", "polish text with too many English quotes", "make AI text undetectable", "humanizer", "humanize Korean", "AI detector bypass for Korean". Follow-ups — "redo just one category", "adjust polish intensity", "change the genre", "only this paragraph", "second pass" — all use this skill too. Handle simple spelling and typo fixes yourself, use the translation skill for translation, and a separate writing skill for rewrites that add or remove content.
Fast structured generation and serving for LLMs with RadixAttention prefix caching. Use for JSON/regex outputs, constrained decoding, agentic workflows with tool calls, or when you need 5× faster inference than vLLM with prefix sharing. Powers 300,000+ GPUs at xAI, AMD, NVIDIA, and LinkedIn.
Accelerate LLM inference using speculative decoding, Medusa multiple heads, and lookahead decoding techniques. Use when optimizing inference speed (1.5-3.6× speedup), reducing latency for real-time applications, or deploying models with limited compute. Covers draft models, tree-based attention, Jacobi iteration, parallel token generation, and production deployment strategies.
Apple Human Interface Guidelines for Apple Vision Pro. Use when building spatial computing apps, implementing eye/hand input, or designing immersive experiences. Triggers on tasks involving visionOS, RealityKit, spatial UI, or mixed reality.
Designs and fixes CI/CD pipelines for Capacitor apps - GitHub Actions, GitLab CI, Fastlane, or Capgo Cloud Build (npx @capgo/cli@latest build request) - covering web build and tests, cap sync, iOS archive/export/upload with App Store Connect API keys, Android AAB signing and Play upload, Capgo live-update uploads, secrets, caching, and build-number bumping. Use when choosing runner images and toolchains (Node 22/24, macos-26 vs xcode-27 runners, JDK 21, Gradle 9 / AGP 9 for Capacitor 9), or when CI fails with errors like "No signing certificate", "requires a provisioning profile", "Unsupported class file major version", "Android Gradle plugin requires Java 17", "The engine node is incompatible", or "Could not find the web assets directory". Do not use for one-off manual store submission (capacitor-app-store), Capgo Cloud Build credential onboarding in depth (capgo-native-builds), Capgo channel/release strategy (capgo-release-workflows), or test authoring (capacitor-testing)
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
Evaluates code generation models across HumanEval, MBPP, MultiPL-E, and 15+ benchmarks with pass@k metrics. Use when benchmarking code models, comparing coding abilities, testing multi-language support, or measuring code generation quality. Industry standard from BigCode Project used by HuggingFace leaderboards.
Merge multiple fine-tuned models using mergekit to combine capabilities without retraining. Use when creating specialized models by blending domain-specific expertise (math + coding + chat), improving performance beyond single models, or experimenting rapidly with model variants. Covers SLERP, TIES-Merging, DARE, Task Arithmetic, linear merging, and production deployment strategies.
GPU-accelerated data curation for LLM training. Supports text/image/video/audio. Features fuzzy deduplication (16× faster), quality filtering (30+ heuristics), semantic deduplication, PII redaction, NSFW detection. Scales across GPUs with RAPIDS. Use for preparing high-quality training datasets, cleaning web data, or deduplicating large corpora.
High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.
Analyze portfolio allocation drift and generate rebalancing trade recommendations across accounts. Considers tax implications, transaction costs, and wash sale rules. Triggers on “rebalance”, “portfolio drift”, “allocation check”, “rebalancing trades”, or “my portfolio is out of balance”.
Configures and uses snipgrapher to generate polished code snippet images, including syntax-highlighted PNGs, SVGs, and WebP exports with custom themes, profiles, and styling options. Use when the user wants to create code screenshots, turn code into shareable images, generate pretty code snippets for docs or social posts, produce syntax-highlighted images from source files, or explicitly mentions snipgrapher. Supports single-file renders, batch jobs, watch mode, and reusable named profiles via the snipgrapher CLI or npx.
Develop and deploy Lakeflow Jobs on Databricks via DABs, Python SDK, or the CLI. Use when creating data engineering jobs with notebooks, Python wheels, SQL, dbt, or pipelines. Invoke BEFORE starting implementation. For compute (serverless, performance mode, classic), also use databricks-serverless
When the user wants to choose or optimize rendering strategy for SEO. Also use when the user mentions "SSR," "SSG," "CSR," "ISR," "static rendering," "dynamic rendering," "server-side rendering," "client-side rendering," "JavaScript rendering," "pre-rendering," "prerender," "content in initial HTML," or "crawler visibility." For crawl issues, use site-crawlability.
Language-independent tokenizer treating text as raw Unicode. Supports BPE and Unigram algorithms. Fast (50k sentences/sec), lightweight (6MB memory), deterministic vocabulary. Used by T5, ALBERT, XLNet, mBART. Train on raw text without pre-tokenization. Use when you need multilingual support, CJK languages, or reproducible tokenization.
Apply SLDS-compliant UI using the correct blueprints, styling hooks, utility classes, and icons. Use when building any UI that needs SLDS, choosing between Lightning Base Components and SLDS Blueprints, applying styling hooks for theming, using utility classes for layout and spacing, or selecting icons. Triggers include "build a modal", "create a form", "data table", "SLDS styling", "style with hooks", "add an icon".
Collect, download, and verify App Store Connect Analytics reports with asc. Use when users need to discover analytics report requests, select report instances by processing date or granularity, download every segment, or verify downloaded files against Apple's size and MD5 metadata before analysis.
Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.
Provides guidance for performing causal interventions on PyTorch models using pyvene's declarative intervention framework. Use when conducting causal tracing, activation patching, interchange intervention training, or testing causal hypotheses about model behavior.
Data framework for building LLM applications with RAG. Specializes in document ingestion (300+ connectors), indexing, and querying. Features vector indices, query engines, agents, and multi-modal support. Use for document Q&A, chatbots, knowledge retrieval, or building RAG pipelines. Best for data-centric LLM applications.
Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.
Provides guidance for training and analyzing Sparse Autoencoders (SAEs) using SAELens to decompose neural network activations into interpretable features. Use when discovering interpretable features, analyzing superposition, or studying monosemantic representations in language models.
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.
Produces human-sounding text that avoids detectable AI writing patterns. Activates on any writing task — tweets, emails, articles, bios, captions, reports, copy, messages, LinkedIn posts, cover letters, README files, or any content where the output must not read as AI-generated. Enforces banned vocabulary, structural variety, punctuation discipline, accuracy rules, and voice calibration. Use when the user says "write," "draft," "rewrite," "make this sound human," "anti-slop," "not AI," or any variation of wanting authentic-sounding output.
Record human-in-the-loop quality judgments for generated images, voice takes, and videos in short-form production. Use this when a person has reviewed an asset and you need structured verdicts, reasons, issue categories, and rerun guidance without turning subjective approval into untracked chat history.
Periodic pass that keeps a project's verification skill and feature map honest: parallel source readers per feature, one live session driving every feature, at most one PR of proven corrections. Use for /maintain-verification-skill or "audit the verify skill".
Comprehensive SAP ABAP CDS (Core Data Services) reference for data modeling, view development, and semantic enrichment. Use when creating CDS views or view entities, defining data models with annotations, working with associations and cardinality, implementing input parameters, using built-in functions, writing CASE expressions, implementing access control with DCL, handling CURR/QUAN data types, troubleshooting CDS errors, querying CDS views from ABAP, or displaying data with SALV IDA. Covers ABAP 7.4+ through ABAP Cloud.
Integrate Circle Gateway to hold a unified USDC balance across multiple blockchains and transfer USDC instantly (<500ms) via permissionless deposit, burn, and mint workflows. Available on 11 EVM chains + Solana (mainnet and testnet), plus Arc (mainnet and testnet). Use when: enabling chain-agnostic user experiences, low-latency or instant next-block finality is required, capital needs to be pooled across chains for greater capital efficiency, or building apps with consolidated crosschain balances. Triggers on: Gateway, Gateway Wallet, Gateway Minter, gatewayMint, burn intent, unified balance, instant crosschain transfer.
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.
Newsletter curation with content sourcing, editorial structure, and subscriber growth strategies. Covers issue formatting, link roundups, commentary style, and sending cadence. Use for: email newsletters, link roundups, weekly digests, curated content, creator newsletters. Triggers: newsletter, email newsletter, newsletter curation, weekly digest, link roundup, curated newsletter, newsletter writing, newsletter format, subscriber growth, newsletter strategy, content curation, newsletter template
Choose and implement the right Circle wallet type for your application. Compares developer-controlled, user-controlled, and modular (passkey) wallets across custody model, key management, account types, blockchain support, and use cases. Use whenever blockchain wallet integrations are required for onchain application development. Triggers on: which wallet, choose wallet, wallet comparison, EOA vs SCA vs Modular Wallet, custody model, programmable wallets.
Guides the agent through migrating an existing Capacitor app project from CocoaPods to Swift Package Manager (SPM) for iOS dependency management. Covers prerequisite checks, inventorying installed Capacitor plugins, backing up customized iOS project files (Info.plist, AppDelegate.swift, Assets.xcassets, Base.lproj, App.entitlements, GoogleService-Info.plist, .xcconfig files, signing configuration), deleting the existing `ios/` folder, re-scaffolding with `npx cap add ios --packagemanager SPM`, restoring preserved files, re-syncing plugins, and verifying the build. Performs all migration steps manually — does not use the interactive `npx cap spm-migration-assistant` command. Do not use for Capacitor plugin projects, app projects already on SPM, app projects without an existing `ios/` folder, or non-Capacitor mobile frameworks
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.