Prompt Policy Firewall

Scan prompts for policy violations (PII, secrets, injection patterns) before sending data to AI models.

0

Risk Score

Allow with caution

Decision

0

High

0

Medium

0

Low

Findings

No violations found.

Redacted prompt

Findings JSON

How to use Prompt Policy Firewall

Screen prompts for PII, secrets, and injection phrases before sending text to AI APIs.

  1. Step 1: Paste prompt content

    Include the exact text block that would be sent to your model endpoint.

  2. Step 2: Enable the right severity checks

    Keep high severity enabled by default and tune medium or low checks as needed.

  3. Step 3: Review findings and line numbers

    Inspect each hit, confirm if it is real risk, and apply recommendation guidance.

  4. Step 4: Copy redacted output

    Use the redacted prompt version for safer downstream model calls.

Pro Tips

  • Treat leaked keys and private key blocks as compromised and rotate them.
  • Run this as a final safety gate after prompt quality edits.
  • Use findings JSON for audit logs in internal QA workflows.

About This Tool

Prompt Policy Firewall helps enforce prompt hygiene before model calls by flagging secrets, PII, and injection-style patterns locally in your browser.

Frequently Asked Questions

Is this perfect detection?

No. Pattern-based checks catch common risks but can miss obfuscated values.

Can false positives happen?

Yes. Always review flagged lines before deleting data.

Is prompt data uploaded?

No. Scanning runs fully client-side.