---
title: "Detect prompt injection attacks in user-supplied input."
method: POST
path: "/v1/scan/injection"
tags: ["security"]
---

# Detect prompt injection attacks in user-supplied input.

`POST /v1/scan/injection`

## Request

- Base URL: `https://api.agent-toolbox.ai`
- URL: `https://api.agent-toolbox.ai/v1/scan/injection`
- Auth: one of:
  - HTTP bearer
  - none

Cost: 1 credit(s). Free tier: 10 calls per IP (no auth).

## Request body

- ScanInjectionRequest
  - `input` string, required
  - `context` string

## Response `200`

Success.

- ScanInjectionResponse
  - `risk` 'safe' | 'suspicious' | 'injection', required
  - `score` number, required
  - `patterns` string[], required
  - `advice` string, required
  - `context` string

## Other responses

- `400`: Invalid request body.
- `402`: Free tier exhausted or insufficient credits.

---

[API](https://skmtc.dev/agent-toolbox/apis/agent-toolbox-ai-api.md) · [All operations](https://skmtc.dev/agent-toolbox/apis/agent-toolbox-ai-api/llms.txt) · [OpenAPI document](https://skmtc.dev/agent-toolbox/apis/agent-toolbox-ai-api/revisions/07132043190b?raw)