Protection of outgoing messages from AI agents/bots (anti-leak + anti-b

A deterministic "guardian" for weekend messages from your AI agents, Telegram bots, and auto-posters. Two layers WITHOUT LLM (thus reliable, without hallucinations): (1) EGRESS filter — catches leaks of secrets/API keys, private (by login) links, and tokens in the text BEFORE sending; (2) BURST kill-switch — counts the sending rate and automatically pauses during spikes to avoid getting banned/flooded. Plus, a watchdog "guardian behind the guardian": checks every half hour that the protection is not breached or disabled. Fail-open on its own error — does not disrupt your processes.