Generate detailed prompts for AI coding tools (Copilot, Cursor, ChatGPT) from your project's context. It supports two workflows:
Error detection and prompt generation
New project analysis and prompt generation
Features
Commands:
"Prompt Craft: Generate Error Prompt"
"Prompt Craft: Analyze Project and Generate Prompts"
Scans Problems (Diagnostics) to identify errors and build a context-rich prompt
Analyzes project structure, dependencies (from package.json), and README to create 3–5 starting prompts
Outputs to a dedicated Output Channel and auto-copies to clipboard
Optional LLM enhancement via Groq (model llama3-8b-8192) to condense context and generate varied tasks
Quick Start
Open this folder in VS Code.
Install dependencies:
Run: npm install (devDependencies include TypeScript typings)
Build the extension:
Run: npm run compile
Launch the extension in a new Extension Development Host:
Press F5 (or use the provided launch configuration)
Optional: Enable Groq LLM enhancement
Get an API key from https://groq.com.
In VS Code Settings → search for "Prompt Craft" → set promptCraft.groqApiKey.
You can also adjust promptCraft.requestTimeoutMs (default 5000).
If no key is set or the request fails, the extension falls back to template-based generation.
Commands
Generate Error Prompt: Attempts to read the first error from VS Code Problems. If none found, you'll be prompted to paste error output. It then extracts file/line (when available), grabs a code snippet, and generates a detailed prompt.
Analyze Project and Generate Prompts: Reads package.json, README.md (if present), and scans up to 100 files to infer a simple architecture summary. It then emits 3–6 tailored prompts. The output now uses a shared "Project Context" header to avoid repetition and includes domain inference (e.g., e-commerce) and simple controller/model summaries.
Notes and Limitations
Terminal output capture is not part of the stable API. This prototype reads Diagnostics and accepts pasted error output. TODO: Explore optional proposed APIs or task events for richer automation.
Scope focuses on JavaScript/TypeScript Node.js projects to keep the prototype simple and fast. Domain inference and code summaries use simple regexes (no heavy parsers) and may miss complex patterns.
LLM is optional. If disabled/unavailable, prompts fall back to templates.
Development
Source in src/ (TypeScript), compiled to out/.
Key files:
src/extension.ts: Activation, commands, and UI
src/analyzer.ts: Project scanning and heuristics
src/templates.ts: Prompt templates
src/domain.ts: Domain inference and simple controller/model extractors
