Public tool listing

Prompt Injection Scanner details

Scan prompts for jailbreaks, instruction overrides, and hidden payloads. This listing defines the commercial catalog page, trust model, and developer handoff for prompt security workflows.

AI securityFree trialSecurityPromptOWASP

Open workspace

Overview

Prompt Injection Scanner helps teams review system prompts, tool instructions, and retrieved text before shipping agentic flows. The first pass is rule-based; deeper AI review remains consent-gated.

How it works

1Add prompt surfacePaste a system prompt, tool instruction, user prompt, or retrieved document excerpt.Local

2Run rule scanDetect override attempts, hidden prompt extraction, URL callbacks, and data exfiltration patterns.Scan

3Generate findingsGroup severity, evidence snippets, and suggested guardrails into a review report.Report

4Harden workflowConvert findings into red-team variants and a release checklist.Workflow

Privacy and review model

Local rulesBaseline scan should run without sending prompt content to a model provider.

ConsentAI deep review must show what text will be sent and which model route will process it.

AuditTeam releases should save reviewer, severity, and remediation state.

Implementation handoff

UIWorkspace anatomyProfile sidebar, prompt input, risk report, meter, findings, and guardrail output.Stable

APIScanner contractReturn severity, category, evidence range, remediation, and consent requirement.Next