Privacy — text is processed locally; detection is not a guarantee.
PromptSafe performs redaction locally in this browser tab; it does not send pasted text to a PromptSafe server or save it in browser storage. The app uses bundled static CSS and JavaScript, with no third-party runtime scripts. If you use the hosted copy, the hosting provider still receives the page request and may keep standard access logs. Regex detection is incomplete, so review the output before sharing. Verify the no-egress processing claim in Developer Tools → Network after the page has loaded.
Raw Input
0 chars · 0 words · 0 linesProcessed locally in this tab. The detector can miss secrets; review before sharing.
Sanitized Output
Redactions are highlighted in colour — the copied text contains the plain [REDACTED_*] tokens.
Strictness
0 / 0 rules activeToken Compression
reduce token count before sending to AILossy log cleanup: repeated lines, timestamps, and log levels may be removed. Estimates use characters ÷ 4, not a model tokenizer. Review the output before using it.
Prompt Injection Detection
flag potential prompt injection patternsReversible Redaction
generate a mapping key to restore redacted text locallyThe mapping key contains the original secret values. Treat it like a password. Copy Key writes those originals to your system clipboard; Clear All removes the in-page copy.