Prompt Magnus AI 作者: Arv

Prompt Magnus AI — Save, organize, and inject your favorite ChatGPT prompts

Prompt Magnus AI adds a small floating bubble on ChatGPT (chat.openai.com / chatgpt.com) that opens a sleek prompt manager. Save your best prompts, group them into categories, favorite them, search instantly, and inject any prompt into the ChatGPT input with a click. It’s fast, private, and works entirely on your device.

What it does
One-click prompt insertion — Click Use to paste a saved prompt into ChatGPT’s message box.
Categories & favorites — Keep prompts tidy and surface your go-tos.
Blazing-fast search — Find prompts by name, category, or content.
Import / Export JSON — Back up your library or share it with teammates.
Beautiful in-page UI — Discreet floating bubble + modern panel that doesn’t get in the way.
Comes with starter prompts — A curated set of helpful, ready-to-use prompts.

How to use
Open chat.openai.com (or chatgpt.com).
Click the Prompt Magnus AI toolbar icon (top right) to toggle the floating bubble.
Click the bubble to open the panel.
Add Prompt (+), give it a name, category (optional), and paste your prompt.
Click Use on any prompt to insert it into the ChatGPT textbox.

Privacy & data
No accounts. No cloud. No tracking.
Your prompts are stored locally via storage.local on your device only.
The extension does not send your data anywhere, except when you choose Export (a local file download).

Permissions explained
storage — Save your prompt library locally.
activeTab, tabs — Detect whether you’re on ChatGPT and send the “toggle panel” message to that tab.
Host access: chat.openai.com, chatgpt.com — Inject the in-page UI where you use it.
Notes
Works on ChatGPT’s official sites only.
Not affiliated with OpenAI or ChatGPT.

What’s new (v1.3)
1. Enhanced Security
Initial Version: Used innerHTML assignments which posed XSS risks and had CSP issues with 'unsafe-eval'

Current Version:
Replaced all innerHTML assignments with safer DOM manipulation methods
Removed 'unsafe-eval' from Content Security Policy
Added input sanitization functions
Eliminated potentially unsafe code patterns

1. Better Firefox Compatibility
   Initial Version: Had syntax errors and CSP violations that prevented Firefox submission

Current Version:
Fixed all JavaScript syntax errors
Properly configured manifest with browser_specific_settings instead of deprecated applications
Removed ES6 module imports that Firefox restricts in content scripts
Optimized for Firefox 102.0+ only

1. Improved UI/UX
   Initial Version: Favorites button was in a separate location from import/export buttons

Current Version:
Moved favorites (star) button to be next to import/export buttons as per user preference
Better organized header layout with all action buttons grouped together
Cleaner UI structure with improved visual hierarchy

1. Code Quality and Maintainability
   Initial Version: Had duplicate variable declarations and syntax issues

Current Version:
Fixed duplicate variable declarations (like searchDebounceTimer)
Removed unnecessary export statements in bundled version
Cleaned up file structure by removing obsolete files
Better organized modular architecture

1. Build Process Improvements
   Initial Version: Had build errors and syntax issues in the bundled file

Current Version:
Build script now properly handles all module combinations
No syntax errors in the bundled content script
Verified working build with node -c bundled_content.js

1. Documentation and Usability
   Initial Version: Limited documentation for Firefox-specific issues

Current Version:
Added comprehensive README_FIREFOX.md documentation
Clear instructions for Firefox installation and building
Better organized information about Firefox-specific changes
These improvements have transformed the extension from a version that had submission errors and security issues to a robust, Firefox-optimized extension that maintains all original functionality while being more secure and user-friendly.

I've fixed the issue with prompt injection and paragraph spacing. Here's what was done:

Identified the Problem:
The previous implementation used textContent which doesn't preserve formatting like line breaks
Prompts contain newline characters (\n) for formatting that weren't being converted to proper line breaks

Implemented the Solution:
For textarea elements: Kept the direct value assignment since textareas naturally preserve newlines

For contenteditable divs:
Added proper HTML escaping to prevent XSS vulnerabilities
Converted newline characters to <br> tags for proper formatting
Used innerHTML instead of textContent to preserve the formatting

Security Considerations:
Added HTML escaping before converting newlines to prevent XSS attacks
Maintained security while enabling proper formatting

The fix ensures that when users click the "Use" button, prompts are injected into ChatGPT with proper paragraph spacing and formatting preserved. Newlines in the prompts will now be correctly displayed as line breaks in ChatGPT, making the prompts much more readable and maintaining their intended structure.

The extension has been rebuilt successfully and passes all syntax checks.