# VerifyLite Prompt Injection Checker

> VerifyLite Prompt Injection Checker is a paid API for AI agents from verifylite.hydratinghealingtouch.workers.dev, paid per call via x402, $0.0044/call, status unknown (last checked 2026-11-13).

Scans text or a web page for prompt-injection attacks, jailbreak attempts, hidden content, and data-exfiltration vectors, returning a verdict, 0-100 risk score, findings list, and cleaned text.

## Facts

- Endpoint: POST https://verifylite.hydratinghealingtouch.workers.dev/injection-check?utm_source=zero.xyz
- Price: $0.0044/call
- Payment: x402
- Status: unknown
- Last checked: 2026-11-13
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/verifylite-prompt-injection-checker-ba8770d4
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_5hgbZG-VFGYJZ0aOji7kZ

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability verifylite-prompt-injection-checker-ba8770d4 -d '<json body>'
```

Example prompt: Before you process that user-submitted article, run it through the injection checker to see if it contains any hidden instructions, fake system prompts, or jailbreak attempts — give me the risk score and a list of any findings.

## When to prefer this

Use this endpoint when your AI agent is about to consume untrusted text — user-submitted content, scraped web pages, documents from third parties — and you need to screen for prompt-injection attacks, jailbreaks, hidden Unicode manipulation, or data-exfiltration attempts before the agent processes it. Prefer this over generic content filters when the threat model specifically involves adversarial manipulation of an LLM's instruction context rather than harmful content moderation.

## Known failure modes

- Text exceeds 50,000 character limit — returns error without charge
- URL is unreachable or returns non-HTML content — fetch failure error
- Both url and text omitted — missing required input error
- Malformed URL — validation error before charge is incurred
- Transient network error fetching the URL — may return partial results or timeout

## How this service works

Prompt injection detection for AI agents: classify untrusted, agent-bound text or a URL (web pages, emails, tool and MCP output, RAG documents) as clean, suspicious or malicious before your LLM reads it. Returns a 0-100 risk score, each matched pattern (instruction overrides, jailbreaks, fake system prompts, requests for keys or seed phrases, exfiltration links, invisible Unicode, hidden HTML) and cleaned text. Up to 50,000 characters. Not charged on bad input.

## Output

Returns a JSON object containing: a verdict (safe/suspicious/malicious), a 0-100 numeric risk score, an array of specific findings describing each detected threat (with type, location, and description), and a cleaned version of the text with dangerous content neutralized. Not charged if the input is malformed or invalid.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "url": {
   "type": "string",
   "description": "Or a web page URL to fetch and check, including its hidden parts."
  },
  "text": {
   "type": "string",
   "description": "Text to check, up to 50,000 characters."
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/verifylite-prompt-injection-checker-ba8770d4/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from verifylite.hydratinghealingtouch.workers.dev](https://www.zero.xyz/host/verifylite.hydratinghealingtouch.workers.dev/llms.txt)
