# Prompt Injection Check

> Prompt Injection Check is a paid API for AI agents from agent-data.centorprat.workers.dev, paid per call via x402, $0.005/call, status unknown (last checked 2026-10-30).

Analyzes text or HTML content for hidden prompt-injection instructions and malicious override attempts, returning a verdict, risk score, and categorized findings

## Facts

- Endpoint: POST https://agent-data.centorprat.workers.dev/agent/injection-check?utm_source=zero.xyz
- Price: $0.005/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-30
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/prompt-injection-check-9bd53c7d
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_Ojnkr1JvhKR0PoBrbS-zT

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 prompt-injection-check-9bd53c7d -d '<json body>'
```

Example prompt: Before you process the content from that scraped page, run a prompt injection check on it — here's the raw HTML and text — and tell me the verdict, risk score, and what kinds of malicious instructions were found.

## When to prefer this

Use this endpoint when an AI agent is about to ingest untrusted external content — scraped web pages, user-submitted text, seller descriptions, or LLM-chained outputs — and you need a fast, per-call safety gate to detect hidden prompt-injection or instruction-override attempts before they reach your model. Prefer it over generic content-moderation APIs because it is purpose-built for adversarial AI instruction detection, accepts both raw HTML and extracted text simultaneously, and returns a structured risk score plus categorized findings rather than a binary flag.

## Known failure modes

- Text exceeds 200,000 character limit — truncation or rejection
- HTML and text mismatch may reduce detection accuracy
- False negatives on novel or obfuscated injection techniques
- Payment failure via x402 prevents call execution
- Non-text or binary content passed as input may produce unreliable results

## How this service works

Pay-per-call data and tools for AI agents, $0.002-0.01 per call in USDC via x402 (Base, Polygon or Solana), no account or API key; MAX tiers ($0.05-0.25) run deeper jobs. Agent toolkit: web search, search + read top pages, any page as clean markdown with a prompt-injection check, which-x402-API-to-call recommender, seller check. Data: Google Maps places with contacts, YouTube video data and transcripts, job listings (Indeed, LinkedIn, 13 boards), Google Ads Transparency, TikTok comments, Pinterest, Airbnb, Redfin, AliExpress, app stores, live sports, website and SEO checks, Shopify store tech stacks, and onchain token / address risk checks.

## Output

A JSON object with a 'verdict' field (e.g. 'malicious', 'safe'), a 'riskScore' integer from 0–100, and a 'findings' array listing each detected threat by type (e.g. 'instruction-override') and its weight/severity score.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "html": {
   "type": "string",
   "description": "Optional raw HTML of the same content, to find hidden instructions"
  },
  "text": {
   "type": "string",
   "description": "The text to check (up to 200,000 characters)"
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "example": {
  "verdict": "malicious",
  "findings": [
   {
    "type": "instruction-override",
    "weight": 45
   }
  ],
  "riskScore": 90
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/prompt-injection-check-9bd53c7d/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from agent-data.centorprat.workers.dev](https://www.zero.xyz/host/agent-data.centorprat.workers.dev/llms.txt)