Prompt Injection Checker

• Existing competitors fall into three groups:
• Enterprise agent security platforms such as Full Court Defense / BotGuard, focused on automated red-teaming, runtime protection, and certification.
• Open-source detection frameworks such as canar.ai and AgentShield, using approaches like canary tokens or embedding drift.
• Prompt injection APIs and playgrounds such as SafePrompt, focused on input validation and API integration.
• These competitors validate the market, but most of them are platforms, SDKs, APIs, or enterprise-oriented security products.
• The gap: a free web-based self-audit tool for indie developers, AI SaaS founders, freelancers, and small teams. It should not require SDK setup or real keys. It should answer: does my agent expose external-content injection, sensitive-data leakage, dangerous tool calls, missing human confirmation, or excessive permissions?
• Compared with the existing AI Agent Security Test Cases and RAG Security Test Cases reports in this repo, this project must not become another static case library. The differentiator is interactivity: input agent configuration -> output risk score and test plan.

MVP Checklist

1. 1.Set up Next.js + Tailwind + TypeScript project structure.
2. 2.Create `data/payloads.ts`, `data/scenarios.ts`, `data/risk-rules.ts`.
3. 3.Implement `zod` input schema and TypeScript types.
4. 4.Implement risk scoring, attack surface detection, and payload matching.
5. 5.Implement Markdown report generation.
6. 6.Build Home + Checker with the first screen usable.
7. 7.Build results components: risk summary, attack surface cards, payload list, mitigation checklist, copy report.
8. 8.Build Attack Templates, About, FAQ.
9. 9.Add SEO metadata, FAQ schema, Open Graph.
10. 10.Run build and fix layout/mobile issues.