Score and enhance your AI prompts before sending them. Promptly analyzes your prompts against proven frameworks and suggests project-aware improvements -- so you get better results from Cursor, Copilot, and other AI coding assistants.
Features
Prompt Scoring -- Evaluate prompts on a 0-100 scale across multiple dimensions
Project-Aware Suggestions -- Recommendations reference your actual files, dependencies, patterns, and git history
Auto Enhancement -- One-click prompt improvement with project-specific context injected
Multiple Frameworks -- Choose the scoring model that fits your workflow:
Blocking Mode -- Optionally pause AI submissions for review before sending
LLM-Powered or Heuristic -- Works offline with heuristics, or use Groq/OpenAI/Anthropic for deeper analysis
Quick Start
Install the extension
Open a project in VS Code or Cursor
Press Cmd+Shift+E (Mac) or Ctrl+Shift+E (Windows/Linux) to score a prompt
Review the score and copy the enhanced version
Commands
Command
Description
Promptly: Score Current Prompt
Score a prompt you type in
Promptly: Enhance Prompt
Enhance a prompt and copy to clipboard
Promptly: Open Score Panel
Show the detailed score panel
Promptly: Select Framework
Switch between CICS, DPI, or Active Partner
Settings
Setting
Default
Description
promptly.framework
cics
Scoring framework
promptly.scoreThreshold
60
Show enhancement when score is below this
promptly.autoEnhance
false
Auto-copy enhanced prompt when score is low
promptly.blockingMode
false
Block prompt submission until approved
promptly.llmProvider
heuristic
Scoring engine: heuristic, groq, openai, or anthropic
promptly.groqApiKey
Groq API key (free tier available at groq.com)
promptly.openaiApiKey
OpenAI API key
promptly.anthropicApiKey
Anthropic API key
How Scoring Works
Promptly analyzes your project to build a profile including:
Language and frameworks detected from config files
Directory structure and architectural patterns
Git history -- recently changed files and hotspots
Import graph from open files
Project conventions from .cursorrules or AGENTS.md
This context powers project-specific suggestions. Instead of generic advice like "Specify which files to modify", you get actionable feedback like "Consider referencing src/auth/login.ts (changed 5 times recently)".
