What problem does it solve? Building multi-turn conversations for language models in Effect requires correctly assembling messages, parts, tool calls, and history without mutating state or mishandling streaming responses. This Skill provides the complete Prompt API reference so agents construct, merge, and serialize prompts correctly on the first attempt. ## Core Features & Use Cases - Message and Part Construction: Create system, user, assistant, and tool messages with text, reasoning, file, tool-call, tool-result, and tool-approval parts using typed constructors. - Prompt Composition and History: Merge prompts with Prompt.concat, manage system messages with setSystem/prependSystem/appendSystem, and convert streaming responses into history via Prompt.fromResponseParts. - Serialization and Provider Options: Encode prompts to JSON with Schema codecs and attach provider-specific options such as OpenAI prompt cache breakpoints or Anthropic cache control. - Use Case: When building a chat interface with LanguageModel.generateText, use this Skill to maintain conversation history in a SubscriptionRef, append user input, and fold model responses back into the prompt. ## Quick Start Use the effect-ai-prompt skill to build a multi-turn conversation with a system message and pass it to LanguageModel.generateText.
