PromptSense

PromptSense is a VS Code extension for .prompt files that provides structured authoring, validation, and intelligent tooling for AI prompt workflows. It treats prompts as structured, analyzable assets by supporting a config + template format.

🚀 Key Features

• Syntax Awareness: Highlight keywords (model, schema, input, etc.) and template variables ({{variable}}).
• Structured Parsing: Automatically detects frontmatter boundaries and template content.
• Intelligent Autocomplete: Suggestions for configuration keys, model names, and schema references.
• Variable Detection: Extracts and tracks template variables across the file.
• Validation: Surface diagnostics for missing fields, invalid configuration, and mismatches between variables and schemas.
• Hover Intelligence: Show schema details and variable context on hover.

📖 Core Concept: The .prompt File

A .prompt file combines structured configuration (YAML-like frontmatter) with a dynamic template.

---
model: gemini-2.0-flash
input:
  schema: UserProfile
---
Hello {{name}}, welcome to PromptSense!

🧬 Schema Support

PromptSense is framework-agnostic and supports multiple schema definition strategies:

1. JSON Schema: Inline or referenced objects.
2. Picoschema: Lightweight schema definitions.
3. Code-defined Schemas: References to schemas defined in your application code (e.g., Gentkit-style ai.defineSchema).

⚙️ Extension Settings

• promptsense.enable: Enable/disable PromptSense features.
• promptsense.models: Custom list of AI models for autocomplete.

🛠️ Development

PromptSense is strictly a development-time tool. It does not execute prompts or call AI models directly.