Hub / LLM Firewalls & Prompt Guardrails / PromptShield

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
Visit PromptShield ↗

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.

Head-to-Head Comparisons

PromptShield vs Lakera GuardPromptShield vs Prompt Security