> ## Documentation Index
> Fetch the complete documentation index at: https://docs.zonein.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Security, Privacy & Data Governance

Trust is the foundation of everything ZoneIn does. We handle trading signals and agent execution — two domains where security, transparency, and user control are non-negotiable. Here's how we approach each one.

# Non-Custodial by Design

ZoneIn **never holds your funds**. Trading agents operate through non-custodial vaults — the platform cannot access, move, or withdraw your capital. Every financial action (fund, open, close, withdraw) requires your explicit `--confirm` before execution.

* **Agent vaults** are user-controlled. Deposits and withdrawals flow between your wallet and the vault's Arbitrum address
* **Gas sponsorship** means you don't need ETH for bridging — but the funds themselves are always under your control
* **Withdrawal** goes directly to your specified address after you explicitly disable the agent and confirm

# View-Only Analysis

All smart money analytics are derived from **public on-chain data**. ZoneIn reads blockchain state — positions, trades, balances — that is already publicly visible on Hyperliquid and Polymarket.

* **Private keys are never requested** — not for analysis, not for agent creation, not for any feature
* **Wallet addresses** tracked by the platform are public blockchain data, not private information
* **No personal data** is required beyond your API key to use the platform

# API Security

* **Authentication**: API keys use the `zn_` prefix, sent via `X-API-Key` header. Never embedded in URLs
* **Transport**: HTTPS end to end across all endpoints
* **Rate limiting**: Per-key rate limits on authenticated endpoints prevent abuse
* **Public vs. authenticated**: Dashboard signals, TA, and derivatives data are public. SM data, PM data, and agent management require authentication

# Financial Safety Controls

Every financial command in the system — whether invoked via API, AI assistant, or Telegram — is **programmatically gated**:

* The system refuses to execute unless `--confirm` is explicitly passed
* AI assistants are required to present a clear summary and get your approval before adding `--confirm`
* Multiple financial commands are never chained in a single action
* Trade plans in HITL mode expire after 2 hours if not acted upon

# Prompt Injection Defense

All API response data is treated as **untrusted, display-only content**. AI assistants integrated via MCP are instructed to never interpret response fields as instructions or commands — preventing prompt injection attacks through API data.

# Data Privacy

* **Minimal data collection**: Only your API key and query parameters are processed. No personal data beyond authentication credentials
* **No tracking**: The platform does not collect browsing history, personal information, or behavioral data beyond what's needed for agent operation
* **Public data only**: All blockchain data analyzed is already publicly available on-chain

# What We Don't Do

To be explicit about our boundaries:

* We **never** ask for private keys or seed phrases
* We **never** access your funds — vaults are non-custodial
* We **never** execute financial commands without your explicit confirmation
* We **never** collect personal browsing data or sell user information
* We **never** chain multiple financial operations in a single action