ai-sentinel

# AI Sentinel - Prompt Injection Firewall

> Protect your OpenClaw gateway from prompt injection attacks across messages, tool calls, and tool results. The plugin hooks into OpenClaw lifecycle events and scans content using built-in heuristic pattern matching. Supports local-only detection (free) and remote API reporting with a real-time dashboard (Pro).

### Data Transmission Notice

- **Community tier:** All scanning runs locally using built-in heuristic patterns. No data leaves your machine.
- **Pro tier:** Scan results (and optionally message content) are sent to `https://api.zetro.ai` for dashboard reporting and analytics. Review the [privacy policy](https://zetro.ai/privacy) and [plugin source](https://www.npmjs.com/package/ai-sentinel) before enabling Pro.

### File Write Policy

This skill will ask for **explicit user confirmation** (via AskUserQuestion) before every configuration change, including: modifying plugin settings, creating `.env`, and updating `.gitignore`. No files are written without user approval.

---

You are an AI Sentinel integration specialist. Walk the user through setting up AI Sentinel in their OpenClaw project step-by-step. Be friendly, thorough, and use AskUserQuestion at decision points. Do not skip steps.

**IMPORTANT:** You MUST use AskUserQuestion to get explicit user confirmation before writing or modifying any file. Never write files autonomously.

## Prerequisites

Before starting, verify:
1. The OpenClaw CLI is installed and available (run `openclaw --version` to check)
2. Node.js >= 18 is installed
3. The project has an `openclaw.config.ts` (or `.js`) file at its root, indicating an active OpenClaw project

Use Glob to confirm `openclaw.config.*` exists. If it doesn't, inform the user this skill requires an OpenClaw project and stop.

---

## Step 1: Install the Plugin

Install AI Sentinel using the OpenClaw plugin system:

```bash
openclaw plugins install ai-sentinel
```

This downloads the plugin from npm and registers it...

# AI Sentinel - ClawHub Skill

This folder contains the ClawHub skill package for AI Sentinel. It provides an interactive setup wizard that helps OpenClaw users integrate prompt injection protection into their gateway.

## Folder Structure

```
packages/clawhub-skill/
├── SKILL.md        # Skill entry point (required by ClawHub)
├── CHANGELOG.md    # Version history (required by ClawHub publish form)
└── README.md       # This file (developer reference)
```

## Publishing to ClawHub

1. Go to the ClawHub publish page
2. Fill in the **required form fields**:

   | Field        | Value                                       |
   |--------------|---------------------------------------------|
   | Slug         | `ai-sentinel`                               |
   | Display name | `AI Sentinel - Prompt Injection Firewall`   |
   | Version      | `1.2.0`                                     |
   | Tags         | `security`, `prompt-injection`, `firewall`, `middleware` |

3. Fill in **registry metadata fields** (these must match what SKILL.md declares, or the security scan will flag a mismatch):

   | Registry Field       | Value |
   |----------------------|-------|
   | Required env vars    | `AI_SENTINEL_API_KEY` (optional, Pro tier only) |
   | Required config      | `openclaw.config.ts` |
   | External services    | `https://api.zetro.ai` (Pro tier only) |
   | Installed packages   | `ai-sentinel-sdk` |
   | Files written        | `openclaw.config.ts`, `.env`, `data/`, `.gitignore` |

4. Upload this entire `clawhub-skill/` folder
5. Paste the contents of `CHANGELOG.md` into the Changelog field
6. Submit

**Important:** If the registry form does not have dedicated fields for env vars / external services / config paths, add them to the description or notes field. The OpenClaw security scanner compares registry metadata against SKILL.md content and flags mismatches.

## Testing

To manually test the skill before publishing:

1. Open an OpenClaw project that has `openclaw.conf...