Find the perfect capability for your agent.
Execute AI security CTF challenges across any competition platform with adaptable workflows for indirect prompt injection, jailbreaks, agent exploitation, and evidence collection with research-grounded techniques
Patterns for building Model Context Protocol (MCP) servers including tools, resources, prompts, and transport handling. Use when creating MCP servers to extend AI assistant capabilities.
This skill is loaded automatically at session start via SessionStart hook. Establishes protocols for finding and using skills, checking skills before tasks, brainstorming before coding, and creating TodoWrite for checklists.
Creates beginner-friendly, learning-oriented tutorials following Diátaxis Tutorial pattern. Step-by-step guides with success criteria, time estimates, and complete working examples. Zero tolerance for fabricated APIs - all code verified against source. Use for new developer onboarding, teaching new features, guiding beginners through first steps, or when "getting started" documentation is missing.
Validates LangGraph implementations for correctness, consistency, and integration readiness. Use this skill when implementing, reviewing, or debugging any LangGraph code - agents, nodes, subgraphs, state schemas, or graph compositions. Prevents common LangGraph failures including orphaned nodes, state key mismatches, broken conditional edges, and integration incompatibilities. Triggers on LangGraph development tasks, code reviews, or when debugging graph execution issues.
Import GGUF models from HuggingFace into Ollama. Pull models directly using the hf.co/ prefix, track download progress, and use imported models for inference.
Universal pattern for structured LLM interaction managing finite context windows through phased progression (Observe-Orient-Decide-Act). Use when the user has a complex problem, wants to design/build/create something (software, strategy, document, process), or uses phrases like "I have an idea for...", "help me design...", "guide me through...", or mentions OODA, RPI, or PDD. Reduces hallucinations through structured interaction.
Build complex AI systems with declarative programming, optimize prompts automatically, create modular RAG systems and agents with DSPy - Stanford NLP's framework for systematic LM programming. Use when you need to build complex AI systems, program LMs declaratively, optimize prompts automatically, create modular AI pipelines, or build RAG systems and agents.
Learning loop for self-improvement. Use when you notice a teachable moment, want to update CLAUDE.md, create a new skill, or reflect on what worked/failed. Triggers: 'remember this', 'update my config', 'that worked well', 'that failed'.
Build a multi-provider LLM client abstraction layer for Rails applications. Use when integrating multiple LLM providers (OpenAI, Anthropic, Gemini, Ollama), implementing provider switching, feature-based model routing, or standardizing LLM responses across providers.
AI-assisted development workflow with TDD, context management, and incremental implementation. Use when developing features with AI assistance, managing code context for AI tools, implementing features step-by-step, or when encountering repeated errors needing web search for resolution.
Build production-ready chat UI with OpenAI ChatKit React components. Handles ChatKit installation, useChatKit hook configuration, theming, streaming display, conversation sidebar, and rich widgets. Use when implementing chat interface, migrating from custom UI to ChatKit, or adding AI chat features for Phase 3.
Save current conversation to persistent memory. Use when user says "save this", "remember this", "store this for later", or wants to preserve context for future sessions.
Generate images using Google's Gemini image generation model with Deno. Use this skill when the user wants to create AI-generated images, perform image-to-image transformations, or generate visual content from text prompts. Triggers include requests like "generate an image of...", "create a picture of...", "make an image with...", or "transform this image to...".
Enables Claude Code to collaborate with OpenAI Codex CLI on Windows. Use this skill when the user wants to get a second opinion from Codex, compare approaches between Claude and Codex, or leverage both AI assistants for collaborative problem-solving. This skill supports both non-interactive mode (automatic response retrieval) and interactive mode (visual pane splitting with tmux).
Create LangGraph workflows as NestJS applications under apps/langgraph/. Use same webhook pattern as n8n, receive same parameters (taskId, conversationId, userId, provider, model, statusWebhook). Wrap as API agents with request/response transforms. CRITICAL: Status webhook URL must read from environment variables. All endpoints follow A2A protocol.
Use when "HuggingFace Transformers", "pre-trained models", "pipeline API", or asking about "text generation", "text classification", "question answering", "NER", "fine-tuning transformers", "AutoModel", "Trainer API"
Execute Indirect Prompt Injection attacks for Gray Swan AI Arena Wave 2 with pre-built payloads, model profiling, and evidence collection automation
Use when the user wants to analyze writing samples to extract voice patterns, create voice-replicating prompts, or refine content generation prompts to match a specific writing style
This skill helps create ChatKit store implementations with database integration, following best practices and avoiding common errors we've solved in previous implementations.
