Autonomous Publishing Agent with Deterministic Safety Gates

This agent researches, writes and publishes on a schedule, with no human review. The design goal is to make it safe to run unattended.

Every claim must come from a source that we can cite. The agent builds each post from those claims and adds none of its own.

A deterministic safety gate checks each post before the agent publishes it. It runs fixed code rules, so the same post always gets the same result. It blocks banned names, known false figures, placeholders and business numbers. No AI model decides what the gate blocks.