Media
3 items
Description
Detect and redact sensitive data locally before sending prompts to AI. PromptShield helps people avoid pasting sensitive information into AI chat tools by checking prompt drafts in the browser before they are sent. It detects common leak patterns such as email addresses, phone numbers, API keys, internal URLs, and financial amounts. When a risky prompt is found, PromptShield shows a small warning panel and offers a one-click redaction flow. The extension is local-first. Prompt text is scanned in the browser and is not uploaded to PromptShield servers. Local event history stores only risk metadata, finding types, and page URL information in Chrome local storage so users can review recent checks. Supported sites in this release: ChatGPT Claude Gemini Perplexity Perplexity uses a safe fallback: when direct editor replacement is not reliable, PromptShield shows a redacted preview and lets the user select the sanitized text for manual replacement.