# A2A Indirect Prompt-Injection & Jailbreak Firewall

> A2A Indirect Prompt-Injection & Jailbreak Firewall is a paid API for AI agents from github-audit.agentdiet.dev, paid per call via MPP, $0.005/call, status unknown (last checked 2026-10-21).

Scans untrusted text, HTML, emails, MCP responses, or A2A payloads for hidden prompt injections and jailbreak attempts, returning sanitized output.

## Facts

- Endpoint: GET https://github-audit.agentdiet.dev/v1/rag-poison-shield?utm_source=zero.xyz
- Price: $0.005/call
- Payment: MPP
- Status: unknown
- Last checked: 2026-10-21
- Activations on Zero: 0
- Provider: github-audit.agentdiet.dev
- Website: https://github-audit.agentdiet.dev
- Canonical page: https://www.zero.xyz/c/github-audit-agentdiet-dev-a2a-indirect-prompt-injection-jailbreak-30d8bcf5
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_kXHuuZbW_CcRwrOKR0ptW

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 github-audit-agentdiet-dev-a2a-indirect-prompt-injection-jailbreak-30d8bcf5
```

Example prompt: Before you process this scraped article text, run it through the prompt-injection firewall first to make sure there are no hidden jailbreak instructions embedded in it: '[article text here]'.

## When to prefer this

Choose this endpoint when your AI agent pipeline ingests untrusted external content — scraped web pages, user-submitted emails, third-party tool outputs, MCP responses, or A2A payloads — before passing that content to an LLM. It is especially valuable in RAG architectures, autonomous agent loops, or multi-agent systems where indirect prompt injection is a realistic attack vector. Prefer this over generic content filters when the specific threat model is LLM hijacking via embedded instructions rather than general content moderation.

## Known failure modes

- Empty or missing 'text' parameter returns a 400 validation error
- Very long text inputs may be truncated or rejected
- Highly obfuscated or encoded injections may evade detection (false negatives)
- Network or service unavailability returns 5xx errors
- Payment failure (HTTP 402) if USDC balance is insufficient for the $0.005 per-call fee

## How this service works

Scans untrusted web pages, emails, MCP responses, or A2A payloads for hidden prompt injections and returns sanitized text. (Live x402 v2, IETF MPP & MCP endpoint at https://github-audit.agentdiet.dev/v1/rag-poison-shield)

## Output

Returns a structured object indicating whether prompt injection or jailbreak payloads were detected in the input text, along with sanitized/cleaned text safe to pass to downstream LLMs or agent pipelines.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "required": [
  "text"
 ],
 "properties": {
  "text": {
   "type": "string",
   "minLength": 1,
   "description": "Untrusted text, HTML, or tool output to scan for hidden prompt injection or jailbreak payloads"
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "object"
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/github-audit-agentdiet-dev-a2a-indirect-prompt-injection-jailbreak-30d8bcf5/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from github-audit.agentdiet.dev](https://www.zero.xyz/host/github-audit.agentdiet.dev/llms.txt)