Agentic Framework Test

This skill evaluates technical knowledge of the LlamaIndex framework for document ingestion, indexing, and retrieval in LLM applications. Questions must focus on LlamaIndex core components including index types, node parsers, document chunking, storage context, retrievers, query engines, metadata handling, and embedding integration. Questions should test understanding of supported features, component responsibilities, and framework limitations. Avoid end-to-end application scenarios or business use cases.

This skill assesses understanding of LangChain as a framework for orchestrating LLM workflows using chains, agents, tools, memory, and prompt templates. Questions must focus on chain types, agent execution models, tool invocation, memory implementations, output parsers, and prompt templating mechanisms. Emphasis should be on architectural concepts and API-level behavior rather than real-world application scenarios.

This skill evaluates in-depth knowledge of LangGraph for building stateful, graph-based, and multi-agent LLM systems. Questions must focus on graph components such as nodes, edges, state objects, conditional routing, loops, execution flow, retries, error handling, and human-in-the-loop patterns. The assessment should test how LangGraph enables controlled execution and complex agent coordination, including differences from linear chain-based frameworks. This skill should receive the highest depth and difficulty.

This skill focuses on technical prompt engineering techniques used to guide LLM behavior. Questions must cover zero-shot, one-shot, few-shot prompting, Chain-of-Thought (CoT), ReAct prompting, system vs user prompts, and prompt templates. Avoid creative writing or subjective prompt evaluation. Questions should test understanding of prompt structure and intent.

This skill evaluates understanding of embedding techniques used to represent text as numerical vectors for retrieval and similarity tasks. Questions must cover static vs contextual embeddings, common embedding models, vector dimensionality, similarity metrics, and embedding selection for retrieval systems. Emphasis should be on conceptual and technical differences rather than implementation-specific code.

This skill focuses on Retrieval-Augmented Generation (RAG) architectures and workflows. Questions must cover RAG pipeline stages, chunking strategies, retrievers, rerankers, vector databases, hybrid retrieval, and common failure modes of RAG systems. Questions should test architectural understanding and component roles rather than deployment or business use cases.

This skill assesses knowledge of observability and evaluation tools for LLM and agentic workflows. Questions must focus on tracing, logging, prompt versioning, run analysis, dataset creation, evaluation metrics, and debugging LLM pipelines. Avoid operational or DevOps-level monitoring scenarios.

This skill evaluates understanding of AutoGen for multi-agent conversational systems. Questions must focus on agent roles, message passing, agent coordination, human proxy agents, and group chat execution patterns. Emphasis should be on framework concepts and communication mechanisms rather than task-specific applications.

This skill focuses on CrewAI as a framework for role-based, task-oriented agent orchestration. Questions must cover agents, tasks, crews, execution strategies, tool assignment, and coordination logic. Questions should test conceptual understanding of how CrewAI structures and executes multi-agent workflows.