output-dev-prompt-file

To create structured prompt files for LLM workflows, generate .prompt files inside the prompts directory using YAML frontmatter for provider configuration and Liquid.js for dynamic templating. This standardizes prompt structure and versioning across projects.

Versioning prompt files for Output SDK workflows is best handled by using versioned naming conventions like [email protected]. This approach manages updates safely without breaking existing prompts in the workflow.

YAML frontmatter configures LLM providers in prompt files by defining specific fields such as provider, model, temperature, maxTokens, and providerOptions. This enforces consistent provider settings across all generated prompts.

Yes, you can use Liquid.js templating for dynamic prompt content in .prompt files. This enables template variable usage and dynamic content generation within the standardized YAML frontmatter structure.

You need a specific prompts directory inside your workflow to centralize creation and organization of .prompt files. This placement ensures consistent prompt management and proper integration with Output SDK workflows.

Limitations when standardizing prompt file formats include being restricted to YAML frontmatter and Liquid.js templating conventions, requiring strict adherence to naming and placement rules within the workflow's prompts directory.