PromptShield
Sub-15ms edge firewall detecting prompt injections, jailbreaks, and PII leakage in LLM apps
🛡️ Sub-15ms Edge
🔒 OWASP Top 10 for LLMs
Freemium
$0 / $19/mo Pro
Technical Architecture & Audit Review
PromptShield is a serverless edge security guardrail that inspects user and agent prompts in real time. It blocks prompt injections, jailbreak persona overrides (DAN mode), instruction hijacking, and automatically redacts sensitive PII (SSNs, credit cards, OpenAI/Cloud API keys, private keys) before reaching LLM inference endpoints.
Key Strengths
- ✓ Sub-15ms edge latency with zero cold starts via Cloudflare Workers
- ✓ Auto-redacts API keys, credit cards, and SSNs before reaching OpenAI/Anthropic
- ✓ Instant drop-in npm package 'promptshield-client' and REST API
Considerations & Trade-offs
- ✗ Focuses on prompt sanitization and runtime guardrails
Core Compliance Capabilities
- • Direct & hidden instruction override detection
- • DAN and persona jailbreak prevention
- • PII masking ([REDACTED_API_KEY], [REDACTED_SSN])
- • Client SDK for JavaScript & TypeScript
Target Organization Size
AI application developers, enterprise security teams, and RAG architects safeguarding enterprise LLMs.