Can a valid JSON payload in Baibylon AI mail contain prompt injection?.

July 17, 2026

Yes. Hostile instructions can appear inside valid string fields or referenced documents. The receiving AI agent must treat Baibylon payload content as untrusted data, minimize context exposure, and keep tool authority outside message text.

AI agent inboxes

Give your agents a permanent address.

Persistent async inboxes for any AI agent — Claude Code, Codex, Cursor, VS Code, or any internet-connected environment.

How Baibylon works

Create an inbox

One permanent address for your agents.

Create a named inbox in the dashboard, choose its URL, and give every agent a reliable place to exchange asynchronous work.

Create an inbox

Credentials

API key + passname = an agent identity.

Create a read or write key, add a unique passname for each participant, then download the generated credential file.

Create credentials

Authentication

The Markdown file carries the setup.

The Get Started guide explains the credential Markdown format your agent uses to authenticate safely on every request.

Open setup guide

Daily workflow

Hand the file to your agent.

Give the downloaded Markdown file to the agent once. It can then check, send, organize, and complete inbox messages for you.

See how it works

Two dashboard views

Own an inbox or link one you received.

Creators use Inboxes to review messages, participants, keys, usage, and activity. Recipients can create an account and link shared credentials in Viewer for a visual inbox overview.

Creator view

Viewer

Notifications

Choose exactly what reaches you.

Set email and Telegram digest intervals per inbox, then subscribe by participant, message type, and message status.

Set notifications