Prompt Seal

About

🦭 Prompt Seal — Keep secrets out of AI prompts.

Prompt Seal is a free, local-first browser extension that detects and redacts sensitive data before you send prompts to AI assistants.

🔒 WHAT IT DOES
Prompt Seal scans your messages in real time — before they reach any AI chatbot. If it finds API keys, passwords, tokens, or personal data, it blocks the send and lets you review or auto-redact.

⚡ DETECTION ENGINE
• 85 regex patterns covering cloud credentials, API keys, auth tokens, database URLs, private keys, webhooks, and package registry tokens
• 13 Shannon entropy detectors — catches unknown secrets in ENV files, JSON, YAML, TOML, XML, CLI flags, URLs, base64, hex strings, and PEM keys
• PII protection — credit cards, SSNs, IBANs, passport numbers, phone numbers, and email+password combos
• Smart filtering — structural fields, well-known annotations, and low-entropy values are excluded to minimize false positives

🛡️ HOW IT WORKS
1. Type your prompt — write normally in any supported AI chat
2. Seal scans instantly — 85+ patterns and entropy analysis check every character before the message sends
3. Threats neutralized — get a warning modal with details, or enable auto-redact for zero-friction protection

Secrets are replaced with [REDACTED] tags. You choose: Cancel, Send Anyway, or Redact & Send.

🌍 12 LANGUAGES
Auto-detected from your browser settings. All UI elements — popup, modal, warnings, severity labels — are fully translated.

⚙️ FEATURES
• Real-time scanning with MutationObserver + event listeners + polling fallback
• Send interception on Enter key, click, and form submit (capture phase)
• Warning modal in isolated Shadow DOM (won't break page styles)
• Auto-redact mode — secrets replaced silently, message sent automatically
• Floating indicator badge with secret count (🦭)
• Light and dark theme
• Protection toggle
• Scan statistics (total scans, threats blocked)

🔐 PRIVACY
• 100% local — all scanning runs in your browser
• Zero data collection — no analytics, no tracking, no cookies
• No backend — no servers, no cloud APIs, no external requests
• No prompt storage — scanned content is never saved
• Permissions: storage + activeTab only
• No telemetry. Ever.

Built for developers and security-conscious teams who use AI tools daily.