About Build With AI Assistants and Verify What They Produce

Researching a Topic with an AI Assistant Choosing a Model Tier for a Workload Calling Claude Through Google Cloud Vertex AI Retrieval-Augmented Generation (RAG) Wiring a Coding Agent into GitHub Actions Controlling Output Randomness with Temperature Steering Model Behavior with a System Prompt Context Windows and Token Limits in Large Language Models Routing a Request to the Right Handler Applying Discernment to AI-Generated User Experience Analyzing a Dataset with an AI Assistant Running MCP over the STDIO Transport Recognizing When a Subagent Is the Wrong Tool Extending a Coding Agent with Hooks Handling Tool-Use Blocks and Returning Tool Results Writing Agent Instructions the Agent Actually Follows Using Server-Side Built-In Tools Verifying Agent Work with a Verification Skill Using Extended Thinking for Hard Reasoning Tasks Generating a Test Dataset for a Prompt Evaluation Running MCP over the Streamable HTTP Transport How an Agentic Coding Assistant Works on a Codebase The 4D AI Fluency Framework Getting Better Results from a Chat Assistant Designing a Subagent That Reports Reliably Implementing an MCP Client Persisting Project Rules in a CLAUDE.md File Choosing Between a Skill, a Memory File, a Hook, and a Subagent Sending Images to a Multimodal Model Calling Claude Through Amazon Bedrock Splitting a Skill Across Multiple Files Building a Shareable Output as an Artifact Defining a Tool Schema a Model Can Call Correctly Running Model Calls in Parallel and Merging Results Giving a Language Model Tools to Act On External Systems Sharing a Skill Across a Team Building a Prompt Evaluation Workflow Choosing a Permission Mode for an Agent Run Setting Standing Context with Global Instructions Trusting the Output of an Unsupervised Agent Run Grading Model Output with Code-Based Checks Giving an Agent Direct Access to Your Files and Folders Installing a Coding Agent and Running a First Prompt Sentence Embeddings for Semantic Text Representation in NLP Steerability: Why a Model Follows Some Instructions and Not Others Next-Token Prediction in Large Language Model Text Generation Exposing Read-Only Data as an MCP Resource Packaging Expertise as an Agent Skill Structuring a Prompt with XML Tags Running an Agent on Managed Sandboxed Infrastructure Combining Lexical and Semantic Retrieval in One Pipeline Obtaining and Protecting an API Key Diagnosing a Skill That Never Triggers Designing an Assignment That Requires AI Fluency Tracing the Agent Loop: Act, Observe, Decide Defining an MCP Tool with the Python SDK Applying Discernment to AI-Generated Code Diagnosing an Unexpected AI Output by Its Underlying Cause Recognizing the Capabilities and Limitations of Current AI Cutting Cost and Latency with Prompt Caching Working the Explore, Plan, Code, Commit Loop Organizing Work in a Project with Shared Knowledge Distinguishing MCP Hosts, Clients, and Servers Running a Coding Agent Headless and Unattended Keeping a Human in the Loop on Consequential Decisions Reporting Progress and Logs from a Long-Running MCP Call Chunking and Retrieval Quality as the Ceiling on a Grounded Answer Applying AI Fluency to Course Design and Learning Outcomes Description: Telling an AI System What You Actually Want Running Model-Written Code with a Code Execution Tool Making a Model Cite Its Sources Deciding Between a Fixed Workflow and an Agent Using AI as a Learning Partner Without Outsourcing the Learning Delegating to a Subagent with Its Own Context Window Deciding Whether an MCP Server Should Hold State Assessing AI Fluency in Student Work Disclosing AI Use Transparently Handling Privacy and Data Before Delegating to AI Writing SKILL.md Frontmatter and a Triggering Description The Context Window as a Finite Budget You Spend Running the Description-Discernment Loop Shipping Reusable Instructions as an MCP Prompt Ranking Documents with BM25 Lexical Search Writing Clear and Direct Prompts Making a First Language Model API Call Creating a Subagent for a Specialized Task Running the Delegation-Diligence Loop Distributing Agent Configuration as a Plugin What a Model Knows and Where Its Knowledge Ends Steering a Model with Few-Shot Examples Working Safely When an Agent Touches Real Files Steering a Long Agent Session Back On Track Delegation: Deciding What to Hand to an AI System Diligence: Taking Responsibility for AI-Assisted Work Extracting Answers from PDF Documents Applying the Rules That Make a Prompt Cache Hit Discernment: Evaluating AI Output, Process, and Behavior Scoping Server File Access with MCP Roots Reviewing Code with an AI Assistant Grading Model Output with a Model-Based Grader Getting Structured JSON Output from a Language Model Reading MCP's JSON-RPC Message Types Sustaining a Multi-Turn Conversation with a Stateless API Chaining Model Calls into a Workflow Model Context Protocol as a Standard Way to Hand a Model Tools Connecting External Tools to a Chat Assistant Letting an MCP Server Borrow the Client's Model with Sampling Debugging an MCP Server with the Inspector Streaming a Model Response Token by Token Orchestrating a Multi-Turn Conversation with Tools