Skills
96,981 packaged capabilities and playbooks for agents. Install any of them with one command.
Development
This skill provides comprehensive knowledge for working with the Anthropic Messages API (Claude API). It should be used when integrating Claude models into applications, implementing streaming responses, enabling prompt caching for cost savings, adding tool use (function calling), processing images with vision capabilities, or using extended thinking mode. Use when building chatbots, AI assistants, content generation tools, or any application requiring Claude's language understanding. Covers both server-side implementations (Node.js, Cloudflare Workers, Next.js) and direct API access. Keywords: claude api, anthropic api, messages api, @anthropic-ai/sdk, claude streaming, prompt caching, tool use, vision, extended thinking, claude 3.5 sonnet, claude 3.7 sonnet, claude sonnet 4, function calling, SSE, rate limits, 429 errors
Use this skill when working inside the ClawTeam repository itself: local development, debugging, reviewing, testing, validating multi-agent flows, or checking whether a code change actually works end-to-end. Use the repository bootstrap scripts to standardize the local `clawteam` command and to wire project-local `.agents` and `.claude` skills back to this repository. This skill is repository-development oriented, not a general end-user usage guide.
Complete knowledge domain for Cloudflare Queues - flexible message queue for asynchronous processing and background tasks on Cloudflare Workers. Use when: creating message queues, async processing, background jobs, batch processing, handling retries, configuring dead letter queues, implementing consumer concurrency, or encountering "queue timeout", "batch retry", "message lost", "throughput exceeded", "consumer not scaling" errors. Keywords: cloudflare queues, queues workers, message queue, queue bindings, async processing, background jobs, queue consumer, queue producer, batch processing, dead letter queue, dlq, message retry, queue ack, consumer concurrency, queue backlog, wrangler queues
Production-tested setup for Cloudflare Workers with Hono, Vite, and Static Assets. Use when: creating new Cloudflare Workers projects, setting up Hono routing with Workers, configuring Vite plugin for Workers, adding Static Assets to Workers, deploying with Wrangler, or encountering deployment errors, routing conflicts, or HMR crashes. Prevents 6 documented issues: export syntax errors, Static Assets routing conflicts, scheduled handler errors, HMR race conditions, upload race conditions, and Service Worker format confusion. Keywords: Cloudflare Workers, CF Workers, Hono, wrangler, Vite, Static Assets, @cloudflare/vite-plugin, wrangler.jsonc, ES Module, run_worker_first, SPA fallback, API routes, serverless, edge computing, "Cannot read properties of undefined", "Static Assets 404", "A hanging Promise was canceled", "Handler does not export", deployment fails, routing not working, HMR crashes
Guides the user through creating or pasting a CNB access token (PAT) and persisting it, installing the cnb CLI and configuring git credentials, so the cnb command line and git push/pull to cnb.cool work without a password. Use when setting up the CNB toolchain for the first time, when cnb commands return 401/403, or when git push/pull to cnb.cool asks for authentication.
Write high-signal code comments for humans and coding agents. Use when adding inline comments, docstrings, API comments, or local rationale near code. Triggers on: "write better comments", "comment this", "document why", "agent-friendly comments", "explain this code with comments", "docstring".
Writing maintainable code - readability principles, SOLID patterns applied pragmatically, and the judgment to know when rules should bendUse when "code quality, clean code, readability, naming, SOLID, refactor, code review, best practices, maintainable, how should I structure, clean-code, solid, readability, maintainability, code-review, naming, functions, principles" mentioned.
Defines refined code through the three lenses of constraints, simplicity, and parametricity, with rules for type proofs, no partial calls, explicit signatures, unrepresentable wrong states, hot-path discipline, and style. Use when writing, changing, or reviewing code to a refinement bar, or whenever the user asks for code to follow the rules in code-refinements
Review code through BurntSushi's "honest machinery" lens — asking whether APIs, invariants, error contracts, performance boundaries, and defaults accurately represent reality. Use when reviewing library APIs, patches that change complexity or cost, dependency additions, or when the user asks "is this API honest?", "will this panic?", "is the cost visible?", "does this scale?", or "are the defaults right?". Works across languages; see REFERENCE.md for Rust-specific elaborations. Do NOT use for deletion/anti-bloat review (use code-review-russ-cox) or product-vs-library tradeoffs (use code-review-mitsuhiko).
(NS) Reverse-engineer an existing, legacy, or brownfield codebase into a technology-agnostic business description of what the system actually does — business rules, actors, permissions, hierarchies, state lifecycles, and external integrations, without implementation details. Use this skill whenever the user mentions brownfield, legacy code, modernization, reverse engineering, codebase analysis, extracting business rules from code, engenharia reversa, documentar sistema legado, gerar spec a partir do código, or wants a functional description faithful to the current product to support rewrite, migration, onboarding, or audit — even if they do not name this skill. Do NOT use for simple code reviews, bug fixes, or when the user explicitly wants a technical architecture document.
Integrate or upgrade the Contentsquare Flutter SDK -- Session Replay, screen tracking, transactions, analytics, masking, privacy opt-in/out. Always starts with a mandatory Contentsquare project healthcheck (ask for the data source IDs or resolve the Android and iOS bundle ids, call the healthcheck endpoint, report project health) before any other work. Use when working with Contentsquare, CSQ, CSQ.start, StartConfig, AnalyticsOptions, ProductAnalyticsOptions, CSQNavigatorObserver, CSQNavigatorAutoRouteObserver, auto_route, AutoRoute, GoRouter, go_router, CSQWebViewWrapper, configureProductAnalytics, or migrating from the legacy `Contentsquare()` API to `CSQ()` (v3.x to v4.x) or from 4.1.x to 4.5.1.
Create a library-grade Vue composable that accepts maybe-reactive inputs (MaybeRef / MaybeRefOrGetter) so callers can pass a plain value, ref, or getter. Normalize inputs with toValue()/toRef() inside reactive effects (watch/watchEffect) to keep behavior predictable and reactive. Use this skill when user asks for creating adaptable or reusable composables.
Writes an Architecture Decision Record (ADR) for a choice already made: its context, the decision, and the consequences. TRIGGER when: the user asks to create or write an ADR, document a decision, record why something was chosen, capture an architectural decision, or preserve the reasoning behind a finalized technical choice. SKIP for: decisions that have not been made yet (use `/create-rfc`), implementation planning or breaking work into tasks (use `/forge plan`), or general documentation that is not a decision record.
Use this skill to create and publish changelog announcements for new features, improvements, or bug fixes. This skill handles the complete workflow - creating detailed changelog documentation pages, adding sidebar announcement cards, and ensuring everything follows project standards. Use when the user mentions adding changelog entries, documenting new features, creating release notes, or announcing product updates.
Design command-line interface parameters and UX: arguments, flags, subcommands, help text, output formats, error messages, exit codes, prompts, config/env precedence, and safe/dry-run behavior. Use when you’re designing a CLI spec (before implementation) or refactoring an existing CLI’s surface area for consistency, composability, and discoverability.
Read and write large cuPyNumeric arrays to HDF5 with Legate's parallel, distributed HDF5 I/O (legate.io.hdf5: to_file, from_file, from_file_batched). Use when a developer needs to save a cuPyNumeric array to an .h5/.hdf5 file, load an HDF5 dataset into a distributed cuPyNumeric array, read a large HDF5 dataset in chunks, hand arrays to an HPC pipeline as a single file, or accelerate HDF5 disk I/O with GPUDirect Storage (GDS). Do not use it for Parquet/cuDF/raw-binary or other sharded/custom layouts (see the cupynumeric-parallel-data-load skill), Zarr or object-store/S3 output, .npz or pickled archives, plain h5py without cuPyNumeric, or pure array compute such as FFT, matmul, or reductions.
Pre-migration readiness assessor for porting NumPy to cuPyNumeric. Use BEFORE substantial porting work begins when the user asks whether code will scale on GPU, whether they should migrate to cuPyNumeric, which NumPy patterns transfer cleanly, what must be refactored before porting, or mentions pre-port assessment, scaling analysis, or refactor planning. Inspect the user's source code, look up NumPy usage, cross-reference the cuPyNumeric API support manifest, and distinguish distributed-scaling-friendly patterns from blockers such as unsupported APIs, scalar synchronization, host round-trips, Python/object-heavy control flow, shape/data-dependent branching, and in-place mutation hazards. Produce a verdict of READY, LIGHT REFACTOR, SIGNIFICANT REFACTOR, or NOT RECOMMENDED, with concrete refactor pointers.
Load a sharded, on-disk dataset (sharded .npy, Parquet/Arrow, raw binary, sharded HDF5, custom layouts) into a distributed cuPyNumeric ndarray via a manual partition + leaf @task launch with CPU/OMP/GPU variants. Use when no single-call loader fits, including when per-shard row counts differ across files. Prefer cupynumeric.load or legate.io.hdf5.from_file when they apply.
CSS component library for Tailwind CSS 4 providing pre-built UI components with semantic class names. Use when building web interfaces with HTML/Tailwind that need: (1) Rapid UI development with consistent styling, (2) Accessible component patterns (buttons, forms, modals, etc.), (3) Theme-aware color systems that work across light/dark modes, (4) Responsive layouts with Tailwind utilities. Works with daisyUI v5+ which requires Tailwind CSS v4+.
Use for debugger-driven runtime root-cause analysis in Rider-supported solutions and projects, including .NET/C#, F#, VB, C++, Unity, Unreal Engine, and other GameDev or mixed-language projects. Trigger when runtime state or control-flow evidence is needed — actual values, branch taken, call order, thread context, whether execution reaches a chosen line — and cannot be derived confidently from source or logs; when pinning the relevant path would otherwise mean reading many files or tracing a deep call chain; or on an explicit request to debug, set breakpoints, step, or inspect runtime values. Do not use for compile-time or syntax errors, source-obvious bugs, failures already pinpointed by logs/tests, or code merely named debug/Debugger.
Apply systematic debugging, profiling, and root cause analysis across codebases and technology stacks. Use when investigating bugs, flaky behavior, crashes, stack traces, production issues, memory leaks, performance regressions, or any unexpected behavior where guessing would waste time.
Use when writing or reviewing Python code in this repo that calls Deepgram Conversational STT v2 / Flux (`/v2/listen`) for turn-aware streaming transcription. Covers `client.listen.v2.connect(...)`, Flux models, end-of-turn detection. Use `deepgram-python-speech-to-text` for standard v1 ASR, `deepgram-python-voice-agent` for full-duplex interactive assistants. Triggers include "flux", "v2 listen", "conversational STT", "turn detection", "end of turn", "EOT", "listen.v2", "flux-general-en", "flux-general-multi".
NVIDIA DeepStream SDK 9.0 development with Python pyservicemaker API. Use when building video analytics pipelines, GStreamer-based video processing, TensorRT inference integration, object detection/tracking, or Kafka/message broker integration.
Use this skill to bring any vision model from HuggingFace or NVIDIA NGC into an NVIDIA DeepStream pipeline with end-to-end automation: ONNX download, SafeTensors export, TRT engine build, custom nvinfer bbox parser, multi-stream benchmark, and PDF report. Object detection models only.
Synchronize every user story in a revised design document with one canonical GitHub Issue and milestone. Use for creating, updating, or resuming story issues from a design doc, including feature-delivery. Existing issues are updated in place and changed completed stories are reopened.
This skill MUST be used when the user wants to process a backlog of dev tickets end-to-end with a known project / branch / merge target / task tracker / design reference workflow, but without committing to any specific platform (Jira / Linear / GitHub Issues / Notion / Figma / Zeplin / etc.). Trigger on requests like 「幫我跑任務」、「照我的流程把 backlog 做完」、「review failed 的全部處理掉」、「用我那套開發流程做事」、"work through my task list", "run my dev loop", "process my tickets end-to-end". MUST NOT be used for a single ticket, which belongs to the other dev-workflow skills. MUST ask the user up-front for project root, working branch, merge target, task list location, design reference location, and how to mark a task done — then loops through tickets implementing → verifying → committing → opening PR → merging → syncing → updating the tracker. MUST NOT bake in platform-specific API calls; figure those out at run time from the user's answers.
This skill should be used when the user asks to "write tests", "django tests", "pytest", "test factories", "create test", "add tests", "test coverage", or mentions testing Django applications, fixtures, or factory_boy. Provides pytest-django patterns with factory_boy for test data generation.
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.