JEV REGISTRY • VERIFIED TOOL|SECURITY & GUARDRAILS| 14.6MS EDGE REFLEX
Home/Hub/Edge Prompt Guardrail
Security & Guardrails14.6ms Reflex$0 / Apache 2.0
18,700 active downloads

Edge Prompt Guardrail

Sub-15ms prompt injection firewall and token exfiltration blocker

jevproxy // prompt-guardrail (reflex kernel)
$npx jevproxy run prompt-guardrail
JevProxy Intercept Resolved(Mode: DIRECT_REFLEX)
14.6ms|Cost: $0.0001|0 reasoning tokens burned
STDOUT • Tool Call Output:Status: 200 OK
agent.dispatch(guardrail_check)
payload: { "input_text": { "type": "string", "description": "Raw prompt or output buffer to evaluate" }...
✔ Decision short-circuited in 14.6ms without roundtrip to frontier LLM.
Upstream token bill saved: $0.0142 on this turn.
Edge Prompt Guardrail
Compatible with Cursor, Claude Code, Windsurf, OpenCode
TRADITIONAL LLM CALL:UNOPTIMIZED
• Median Latency: 1310 ms
• Cost per Turn: $0.0142
• Mode: Full KV Cache Reload & TTFT Prefill
JEVPROXY REFLEX KERNEL:77x FASTER
• Median Latency: 14.6 ms
• Cost per Turn: $0.0001
• Accuracy: 99.2% deterministic

1-Click CLI Execution

Run this tool accelerated through the JevProxy gateway without manual wiring:

npx jevproxy run prompt-guardrail

OpenAI / Anthropic Tool Schema

JSON Specification

Paste this schema into your agent tools definition or Cursor extensions:

{
  "type": "function",
  "function": {
    "name": "guardrail_check",
    "description": "Scan text for prompt injection, jailbreaks, and credential leaks",
    "parameters": {
      "type": "object",
      "properties": {
        "input_text": {
          "type": "string",
          "description": "Raw prompt or output buffer to evaluate"
        },
        "strict_mode": {
          "type": "boolean",
          "description": "Enable aggressive secret token filtering"
        }
      },
      "required": [
        "input_text"
      ]
    }
  }
}

Technical Architecture & Usage

Security at Line Speed

Traditional LLM guardrails rely on secondary LLM calls that add 1.2 to 2.0 seconds of latency to every turn. Edge Prompt Guardrail uses calibrated classification heads to detect attacks in 14.6ms.

Accelerate Edge Prompt Guardrail with JevProxy

Get 5,000,000 free decision tokens. Eliminate the 3-second tool freeze in Cursor and Claude Code in under 60 seconds.