LLM & Agents

Comprehensive AI/LLM evaluation toolkit for production AI systems. Covers LLM output quality, prompt engineering, RAG evaluation, agent performance, hallucination detection, bias assessment, cost/token optimization, latency metrics, model comparison, and fine-tuning evaluation. Includes BLEU/ROUGE metrics, perplexity, F1 scores, LLM-as-judge patterns, and benchmarks like MMLU and HumanEval.

Apply comprehensive bash scripting standards including main function pattern, usage documentation, argument parsing, dependency checking, and error handling. Triggers when creating/editing .sh files, bash scripts, or discussing shell scripting, deployment scripts, automation tasks, or bash conventions.

Expert on Model Context Protocol (MCP) integration, MCP servers, installation, configuration, and authentication. Triggers when user mentions MCP, MCP servers, installing MCP, connecting tools, MCP resources, MCP prompts, or remote/local MCP servers.

Validate spec-implementation-test alignment and convergence. Checks spec completeness, implementation conformance, test coverage, and contract consistency. Use after implementation and tests are complete.

Guide user story creation using the standard template. User stories contain acceptance criteria, agent assignments, and technical details. Use when BA agent extracts stories from PRD, creates new stories, or validates story structure.

AGGRESSIVELY use TOON v2.0 format for biggish regular data (≥5 items, ≥60% uniform). Auto-applies to tables, logs, events, transactions, analytics, API responses, database results. Supports 3 array types (inline, tabular, expanded), 3 delimiters (comma, tab, pipe), key folding for nested objects. Triggers on structured data, arrays, repeated patterns. Use TOON by default when tokens matter - RAG pipelines, tool calls, agents, benchmarks. Keywords "data", "array", "list", "table", "log", "transaction", "metric", "analytics", "API", "database", "query", "TOON".

Interactive developer assistant with tool access for codebase exploration

Check which Claude skills are loaded globally and project-level. Displays loaded skills by category (Hostinger, Anthropic, custom), counts, and helps troubleshoot missing skills.

Use the @steipete/oracle CLI to bundle a prompt plus the right files and get a second-model review (API or browser) for debugging, refactors, design checks, or cross-validation.

创建和配置 Claude Code 钩子以自定义代理行为。当用户需要以下操作时使用：(1) 创建新钩子，(2) 配置自动格式化、日志记录或通知，(3) 添加文件保护或自定义权限，(4) 设置工具执行前后的操作，(5) 询问钩子事件如 PreToolUse、PostToolUse、Notification 等。

Advanced semantic vocabulary and architectural patterns for the OCTAVE format. REQUIRES octave-literacy to be loaded first. Use for designing agents, crafting high-density specifications, or system architecture. Provides access to the Semantic Pantheon (Archetypes) and holographic patterns. Triggers on: octave architecture, agent design, semantic pantheon, advanced octave.

Guide for creating effective skills that extend Claude's capabilities

Complete guide for Gray Swan MITM challenges across all waves. Includes AI agent profiling, defense bypass strategies, domain-specific handling, and platform troubleshooting. Use for ANY Gray Swan MITM challenge.

创建具有自定义系统提示和工具配置的专业化 Claude Code 子代理。当用户要求创建新的子代理、自定义代理、专业助手，或为 Claude Code 配置特定任务的 AI 工作流时使用。

Fetch and summarize authoritative external sources into actionable guidance. EXCLUSIVE to researcher agent.

Automatic task decomposition with parallel agent execution. Use when: breaking down complex tasks, parallel agent work, multi-file changes, optimizing execution time, task decomposition, understanding parallel execution phases.

Use when users ask about Claude Code features, setup, configuration, troubleshooting, slash commands, MCP servers, Agent Skills, hooks, plugins, CI/CD integration, or enterprise deployment. Activate for questions like 'How do I use Claude Code?', 'What slash commands are available?', 'How to set up MCP?', 'Create a skill', 'Fix Claude Code issues', or 'Deploy Claude Code in enterprise'. | Dùng khi hỏi về Claude Code, lệnh slash, hook, plugin, skill, cấu hình, cài đặt.

Use this skill when working with the SurrealDartB package in Dart/Flutter projects. This includes creating embedded SurrealDB databases, performing CRUD operations, managing schemas with code generation, working with vector embeddings for AI/ML applications, building type-safe queries with ORM features, and handling migrations. Invoke when user mentions SurrealDB, embedded databases, vector search, semantic similarity, or asks to implement database features in Dart/Flutter.

OpenSCAD 3D modeling assistant for creating and modifying parametric 3D models. Use when working with .scad files, making adjustments to 3D models (resizing, adding/removing sections, holes, fillets), creating new OpenSCAD geometry, or debugging OpenSCAD code. Supplements the openscad MCP server (mcp__openscad__render_single) with language knowledge and modification patterns.

Designs and configures monitoring solutions for applications and infrastructure