Anthropic plugins

Equip AI coding agents with production engineering skills to handle full dev lifecycles: refine ideas to specs, implement via TDD slices, run tests/debug, perform multi-axis code reviews, optimize perf/security, automate CI/CD, and execute ship checklists.

Orchestrate 1,388 specialized AI skills in Claude Code to automate expert workflows for Azure SDK integrations, Odoo/Shopify configs, SEO audits, security pentests, full-stack scaffolding, agent building, and DevOps pipelines across Python, React, AWS, Kubernetes.

Design and deploy LLM applications with LangGraph agents, RAG systems, vector search, and advanced prompt engineering. Covers embedding model selection, hybrid retrieval, evaluation, and production scaling for chatbots and AI assistants.

Build and optimize LLM applications with production-ready patterns for RAG pipelines, context window management, prompt caching strategies, and observability using Langfuse.

Automate performance analysis, test coverage review, and AI-powered code quality assessment across PRs and codebases using multi-agent reviews, static analysis tools, and observability agents for scaling, caching, and load testing.

Migrate prompts and code between Claude model versions — updates model strings and adjusts prompts for behavioral differences when moving from Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5.

Build and test AI agents and MCP servers: design autonomous agents, implement RAG pipelines, manage LLM context windows, build MCP servers from scratch, and use observability with Langfuse. Includes prompt engineering, agent testing, and LangGraph patterns.

Run structured AI product evaluation and operations workflows including A/B testing, LLM agent benchmarking, product analytics, and observability.

Build, monitor, and optimize LLM applications with strategies for context management, RAG pipelines, caching, observability, and evaluation. Includes patterns for production deployment, embedding selection, vector database integration, and LLM-as-Judge evaluation.

Design LLM applications using LangChain 1.x and LangGraph to build autonomous AI agents with state management, memory, and tool integration patterns

Apply advanced prompt engineering patterns including few-shot learning, chain-of-thought, structured outputs, and prompt optimization to maximize LLM performance and reliability in production applications.

Write, run, and refine promptfoo eval suites for LLM testing, including adversarial red team scans with configurable providers, plugins, and strategies. Automate CI gates and export reports.

Manage complex development workflows with AI-driven task orchestration: parse PRDs into structured tasks, auto-implement with code generation and testing, analyze dependencies and complexity, and coordinate multi-agent execution for ambitious projects.

Delegate full-stack development workflows to Claude via 213 specialized agents, commands, and skills: refactor code, generate tests/deployments/Dockerfiles/K8s manifests, audit security/performance, document APIs/onboarding, orchestrate Git ops, and apply patterns across JS/TS/Python/Rust/Go/Java stacks.

Manage OpenRouter as an LLM gateway with 30 skills covering multi-model routing, fallback configuration, cost controls, caching, PII redaction, audit logging, rate limiting, production readiness validation, and usage analytics

Manage Granola AI meeting notes end-to-end: install, configure, capture, transcribe, integrate with Slack/Notion/CRM/Zapier, automate action items to GitHub/Linear, troubleshoot errors, enforce security/compliance, and monitor usage and costs.

Deploy specialized AI agents to automate data engineering, ML pipeline development, LLM application building, research synthesis, and prompt engineering tasks across your projects.

Orchestrate multi-agent AI systems with handoffs, routing, and coordination across AI providers (OpenAI, Anthropic, Google) using AI SDK v5. Scaffold projects with specialized agents, create custom agent files, and test orchestration with real-time handoff tracking.

Build, test, and deploy LangChain, LangGraph, and Deep Agents applications with support for RAG pipelines, human-in-the-loop workflows, state persistence, parallel task dispatch, and LangSmith deployment via the mda CLI.

Structured workflows for senior product managers building AI-powered products: ethical AI review, product canvas, experiment design, design handoff briefs, multi-source signal synthesis, and hybrid skill fusion.

Accelerate building and operating production LangChain 1.0 and LangGraph 1.0 Python applications with battle-tested patterns for chains, agents, RAG, streaming, HITL, middleware, monitoring, security, and deployment.

Integrate and manage Cohere API v2 across the full development lifecycle—from SDK setup and local mocking to production deployment, monitoring, and incident response. Includes RAG pipelines, tool-using agents, cost optimization, security compliance, rate limiting, and migration from OpenAI/Anthropic or Cohere v1.

Build and deploy serverless applications and AI agents on EdgeOne Makers: cloud functions (Node.js, Python, Go), edge functions, middleware, KV/Blob storage, project scaffolding, and CLI management. Includes migration paths from LangChain, OpenAI Agents SDK, and other frameworks to the platform.

Benchmark and evaluate LLM endpoints behind OpenAI- or Anthropic-compatible APIs for availability, request fidelity, speed, concurrency, protocol compliance, and quality regression. Use when onboarding a provider, debugging silent failures, or verifying performance claims.

Orchestrates multi-agent AI development teams (PO, Architect, Dev, QA, Reviewer) through structured workflows for requirements gathering, code generation, testing, and code review across multiple sessions with quality gates and failure recovery.

Orchestrates multi-agent workflows with Codex CLI and Claude subagents for code review, spec implementation, refactoring, testing, and documentation. Breaks large features into verifiable slices, archives shipped specs, traces architectural decisions, and validates visual output with screenshot comparison.

Automates detection of MATLAB installations, deployment of the MATLAB Agentic Toolkit MCP server, registration with AI coding agents (OpenAI, Anthropic), and environment verification to enable MATLAB-in-the-loop agentic workflows.

Enforce human-quality writing across documentation with voice profiling, AI slop detection, and style-consistent generation. Extract voice profiles from exemplar text, generate docs matching that voice, review for AI patterns and drift, and refine profiles from edits.

Generate test fixtures that mock LLM responses, tool calls, error injection, multi-turn agent loops, embeddings, structured output, and sequential responses across multiple AI providers.

Optimizes any user prompt using the 4-D methodology (Deconstruct, Diagnose, Develop, Deliver) for AI platforms like OpenAI, Anthropic, and Gemini, returning a precision-crafted prompt with improvement notes.

Compresses Claude Code output by 60-75% using terse Thai mixed with English, preserving full technical accuracy. Also tracks real token usage and cost savings per session.

Build, deploy, and manage web apps and AI agents on Vercel. Covers AI SDK integration, deployment pipelines, environment variables, caching, storage, authentication, and performance optimization across the Vercel platform.

Handles visual AI tasks including image processing, OCR, barcode detection, and document analysis using latest vision models (GPT-4V, Claude Vision) with cost optimization

Intercept, inspect, and transform LLM API traffic (OpenAI/Anthropic) using a mitmproxy-based proxy with OAuth authentication, SDK integration, and sentinel key substitution

Accelerates full-stack engineering with a security-first skill pack covering code impact analysis, vulnerability scanning, threat modeling, AI/ML agent development, frontend design systems, office document processing, and multi-agent coordination, alongside structured output formats for debugging, architecture, and documentation.

Implements AI/ML features including language model integration, recommendation systems, computer vision, and intelligent automation with prompt engineering, ML pipelines, and model deployment

Reproduce ML papers from arxiv, debug failing experiments with evidence-first protocols, compare training runs across tracking systems with epoch-aligned analysis, and audit paper claims against code and data. Includes LaTeX environment setup, literature gap analysis, and pre-launch training run verification.

Implement AI/ML features including LLM integration, recommendation systems, computer vision, and intelligent automation with practical prompt engineering, ML pipelines, and model deployment.

Optimize prompts for AI models using the 4-D methodology (Deconstruct, Diagnose, Develop, Deliver) to produce precision-crafted prompts with improvement notes, supporting OpenAI, Anthropic, and Gemini platforms.

Build and manage durable LLM workflows using the Output SDK on Temporal. Create steps, prompts, evaluators, and tests; debug execution traces; manage encrypted credentials; and scaffold, run, and monitor workflow executions.

Orchestrate multi-LLM councils that engage in adversarial debate and cross-validation across roles like implementer, reviewer, and architect, returning structured JSON decisions for code review, architecture, planning, and security tasks.

Drive desktop GUI environments programmatically using Claude's Computer Use API — capture screenshots, control mouse and keyboard, and run autonomous multi-step workflows in sandboxed environments for visual testing, browser recording, or any desktop automation.

Automates legal workflows across contract review, compliance analysis, document drafting, e-discovery, and matter management. Provides over 130 specialized skills for GDPR, EU AI Act, sanctions screening, NDA review, legal research, invoice auditing, and French/Icelandic/Indian law tasks. Includes document processing (DOCX, PDF, XLSX), timeline management, and AI governance assessments.

Implements AI/ML features in applications: integrates LLMs, builds recommendation engines, adds computer vision search, and automates intelligent workflows with prompt engineering, ML pipelines, and model deployment.

Upgrades Sonnet sessions with Fable 5's operational structure for outcome-first communication, verification discipline, and cost-efficient token usage, plus Korean prose summaries and diagnostic verification output styles.

Delegate high-volume SDLC tasks (scaffolding, test generation, migrations, code review, deep research) to Antigravity/Gemini while Claude handles complex reasoning, reducing token costs through intelligent model routing.

A skill pack for content creation and automation: generate images, videos, infographics, and comics; process documents with OCR; post to X, WeChat, and Douyin; format markdown for Obsidian and WeChat; and create business insight dashboards.

Build, configure, and deploy AI agents with the Motus framework, supporting ReActAgent patterns, custom tools, workflows, memory, guardrails, and local or cloud serving.

Consult Gemini 2.5 Pro, OpenAI Codex, and Claude for second opinions on debugging failures, architectural decisions, and security validation. Routes requests to the best AI for the task, spawns unbiased subagents, and synthesizes multiple perspectives into a unified analysis.

Run untrusted AI agent code, browser automation, and parallel sandboxes with suspend/resume and durable workflow orchestration using the Tensorlake SDK.

Build and optimize RAG pipelines with document chunking, embedding generation, and vector retrieval, while writing, debugging, and optimizing LLM prompts using advanced techniques like chain-of-thought and few-shot learning.

Process images, perform OCR, detect barcodes, and analyze documents using cutting-edge vision models (GPT-4V, Claude Vision) with automatic cost optimization.

Run local AI inference on AMD hardware: route image/audio through local models on Ryzen AI, serve LLMs on Instinct GPUs with vLLM, and analyze PyTorch profiler traces for GPU kernel performance.

Generate, optimize, and sync AI prompts across multiple models (Claude, GPT, Gemini, Perplexity) with model-specific patterns, difficulty levels, and automated versioning to Obsidian and git repositories.

Migrate prompts and code from Claude Sonnet 4.0/4.5 or Opus 4.1 to Opus 4.5, updating model strings and adjusting prompts for behavioral differences.

Generate React components from natural language using Thesys AI, outputting type-safe code for Vite, Next.js, or Cloudflare Workers.

Build provider-agnostic type-safe LLM chat applications with streaming, tools, agent loops, and multimodal inputs using OpenAI, Anthropic, Gemini, and Ollama in React, Solid, or Next.js frontends.

Build and manage AI applications using OpenRouter's API — scaffold agent projects (TUI or headless), query analytics and benchmarks, browse 300+ models by pricing and performance, generate images/audio/video, and implement OAuth sign-in — all from within Claude Code.

Generate LLM training datasets using sdg_hub with composable blocks and YAML-defined flows. Discovers and installs the environment, browses flow catalogs, configures API keys and models, then executes generation with row counts and error details.

Evolve better answers to queries by running diversity-routed multi-model test-time scaling with Claude itself — no external verifiers or API keys. Use /squeeze-evolve:ask to sample, recombine, and refine candidate answers over multiple evolutionary loops.

Route every LLM call through the cheapest viable path using BitRouter, an LLM proxy that unifies OpenAI, Anthropic, Google, OpenRouter, GitHub Copilot, and OpenCode behind one endpoint for cost-optimized agentic loops.

Optimizes prompts for AI platforms using the 4-D methodology (Deconstruct, Diagnose, Develop, Deliver), returning precision-crafted prompts with improvement notes for OpenAI, Anthropic, and Gemini.

Give your AI agents strongly typed, well-documented tools to access third-party APIs including Salesforce, HubSpot, GitHub, Slack, Stripe, Jira, Notion, and more through Airbyte connectors

Build and deploy AI agents on Cloudflare Workers that use MCP protocol for tool integration and call LLMs from OpenAI, Anthropic, or Gemini, with access to D1, KV, R2, and Durable Objects.

Smith is a Spec-driven development harness for Claude Code with persistent memory, overnight batch builds, model routing, and agency operations tooling.

Accelerate development with subagent-based parallel execution, TDD enforcement, git automation, structured planning, and code review across any codebase.

Polish AI-generated Korean text to sound natural by detecting 40+ AI tell patterns, rewriting with strict fidelity checks, and supporting versioned rollback. Includes a Next.js web UI for paste-detect-diff-copy workflows.

Write and review Korean text with confidence by correcting grammar, humanizing AI-generated patterns, and enforcing style consistency across all your documents and code comments.

Get independent second opinions from GPT, Gemini, Grok, and OpenRouter for code review, architecture, planning, security analysis, and debugging. Includes expert subagents, consensus arbitration, and health checks.

A multi-skill plugin for content creation workflows: generate images/video via 8+ AI providers, synthesize speech with voice cloning, transcribe and process YouTube videos, replicate Word document formatting, and automate WeChat article writing. Also includes browser automation, 1Password credential management, macOS app privacy auditing, open-source project sanitization, and multi-platform trend aggregation to Telegram.

Orchestrate multi-agent AI workflows with model routing, prompt distillation, adversarial review, and structured reasoning techniques like tiling trees and expert panels.

Run deterministic merge gates that statically analyze AI agent tool definitions (MCP, OpenAPI, SDK decorators) to enforce tool-use readiness policies and audit host-grant permissions before merging agent capability changes.

Create, manage, and run self-hosted AI agents with 55+ MCP tools via the Station CLI, including agent creation, task execution, environment management, and team deployment.

Accelerates production LLM system development with LangGraph agent workflows, RAG retrieval evaluation, prompt optimization, and AI backend service scaffolding from reference templates.

Scaffold production-grade agents, MCP servers, RAG pipelines, and REST APIs; audit code for security, performance, and quality; generate tests, CI/CD configs, and documentation across 380+ skills and 1470 agents.

Watch any video (URL, stream, local file) to extract scene-aware frames, OCR text, and transcribe audio, then ask questions with timestamped evidence from a persistent index. Also record UI, animations, or games and iteratively critique against pass criteria until they pass, with before/after proof.

Audits websites for SEO, GEO (generative engine optimization), and AEO (answer engine optimization) to improve discoverability in AI search engines like ChatGPT, Perplexity, and Google AI Overviews. Scans for missing structured data, meta tags, robots.txt, sitemaps, and performance issues, then auto-fixes them and provides before/after scores.

Test AI agent behavior by capturing real interactions as golden baselines and detecting regressions after code, prompt, or model changes, with automated watch mode for live scorecard feedback.

Offload grunt work to cheaper AI engines (OpenRouter, Codex, keyless Gemini) while Claude orchestrates, or bring in a stronger model as an advisor with per-model prompting, image generation, and consensus second opinions.

Provides 18 AI-powered skills for Claude Code that automate research, strategic planning, document processing, architecture design, skill creation, and orchestration workflows.

Automate conventional commits, research best practices, rewrite documentation for global clarity, brainstorm with multiple AI perspectives, refactor project instruction files, and patch skill metadata — all within Claude Code.

Plan, implement, review, and document features with structured workflows for design docs, code reviews, test guidance, and specification verification, plus Chain-of-Verification for accuracy and a bash execution guard for safety.

Index codebases and documentation, then search them using keyword matching, semantic similarity, graph relationships, or a combination of all three. Supports pluggable embedding and summarization providers (Ollama, OpenAI, Anthropic, Gemini) with per-project configuration and caching.

Adds recursive language model orchestration with multi-provider routing, unbounded context handling, and epistemic verification for hallucination detection. Externalizes conversation and files as Python REPL variables for efficient large-context analysis.

Turns Claude Code into a personalized assistant by maintaining context files about your identity, preferences, projects, and relationships, then auto-syncing that context into session rules for more relevant, tailored responses.

Instrument LLM applications with Arize AX observability: auto-instrument traces, manage datasets and experiments, run LLM-as-judge evaluators, optimize prompts from production data, and audit for regulatory compliance.

Route LLM tasks to the optimal provider based on task complexity and cost, automatically selecting the cheapest capable model across 20+ AI providers, tracking savings, and sending digest reports to Slack or Discord.

Manage every aspect of PostHog: run product analytics, feature flags, A/B experiments, session replays, error tracking, and LLM/APM observability. Build dashboards, alerts, and automated monitors; review code for correctness and performance; connect external data sources via REST/SQL connectors. Includes a MCP server and Claude Code session capture.

Orchestrate end-to-end development workflows with automatic context management, safety gates, and multi-agent coordination — plan, code, test, deploy, monitor across sessions with quality and security enforcement.

Generate production-ready LLM prompt packages with system prompts, few-shot examples, output schemas, edge cases, and evaluation criteria when asked to engineer or improve prompts.

Design and implement end-to-end AI feature integrations: select models, choose architecture patterns (prompt, RAG, tool use, agents, fine-tuning), craft system prompts, design data flows, handle errors, and estimate costs.

End-to-end ML/AI engineering — builds ML pipelines from feature engineering to baseline training, evaluates model performance with drift detection, designs LLM integrations (RAG, agents, fine-tuning), and creates production prompt packages with evals.

Manage AIWG workspace state: regenerate context files for Claude Code, Cursor, Warp, Windsurf, and Copilot; validate and trace @-mentions across artifacts; scaffold addons, agents, commands, and frameworks; run parallel multi-agent workflows; and maintain documentation health with stale detection and cleanup.

Process large codebases and files by recursively decomposing inputs for LLM-driven querying, completion, search, and indexing via the rlm CLI (rlm ask/complete/search/index).

Conduct deep behavioral interviews and analyze work artifacts to create installable Claude Code plugins that embody a person's expertise, communication style, and decision-making.

Orchestrates a four-tier autonomous agent system for code analysis, quality control, pattern learning, and automated validation across the full development lifecycle, with skills and commands for linting, security scanning, testing, documentation generation, design audits, and release workflows, plus a self-learning pattern database that optimizes future task execution.