Skip to content

Vivid Labs AI

Now Watching

The Night Kitchen

The Night Kitchen /Agent Capabilities and Prompting Foundations

Agent Capabilities and Prompting Foundations

Establish what makes an AI agent different from a single model response and how prompts shape autonomy, reasoning, tool use, and complex task execution.

Queue

11 videos in this course

The Night Kitchen

Jump across the full chapter sequence without leaving the player.

01

Agent Capabilities and Prompting Foundations

Establish what makes an AI agent different from a single model response and how prompts shape autonomy, reasoning, tool use, and complex task execution.

02

Conversational and Autonomous Agents

Compare conversational assistance with autonomous execution, then design prompts that guide agents toward goals while preserving useful interaction and control.

03

Multi-Step Workflow Design

Break a larger objective into ordered steps and give an agent the context, transitions, and completion criteria needed to execute a workflow reliably.

04

Dynamic Decision-Making and Tool Calls

Prompt an agent to choose its next step dynamically, decide when a tool is warranted, and use observations to continue rather than follow a rigid script.

05

Guardrails, Uncertainty, and High-Risk Domains

Reduce hallucinations with boundaries, uncertainty handling, verification steps, and escalation rules suited to high-impact domains.

06

Aligning Language, Expertise, and Tone

Calibrate terminology, depth, tone, and communication style so an agent responds appropriately to the user's expertise and situation.

07

Lists, Structure, and Output Formats

Specify lists, schemas, field requirements, and formatting constraints that turn free-form model output into predictable information an application can use.

08

Roles, Context, Examples, and Objectives

Combine role, instructions, context, examples, constraints, and success criteria into prompts that stay focused across multi-step tasks.

09

Cost-Aware Prompting and Token Budgets

Estimate token cost, trim unnecessary context, and design prompt strategies that balance reliability with the economics of running agents at scale.

10

Iterative Refinement and Version Tracking

Test, measure, revise, and document prompts so improvements are evidence-based and production behavior remains traceable across versions.

11

Advanced Few-Shot Prompting

Select demonstrations that teach the right pattern, cover meaningful variation, and help an agent generalize to new tasks with fewer examples.