Facio Blog

Practical notes on human-reviewed AI agents.

Payload-powered product notes, security writing, HITL patterns, and operational guidance from the Facio runtime: long sessions, Placet approvals, audit trails, memory, providers, channels, tools, and Docker-first operations.

Showing 31-35 of 81 articles.

Human-in-the-loop

Prompt Injection vs. HITL: Why Your Approval Gates Need Deterministic Enforcement — Not Model Promises

Jun 2, 2026Human-in-the-loop

Prompt Injection vs. HITL: Why Your Approval Gates Need Deterministic Enforcement — Not Model Promises

Prompt injection attackers don't need to hack your infrastructure. They just need to convince your agent to skip the approval gate. Research shows prompt-based HITL fails 74.6% of the time against adversarial input. Here's why — and how deterministic pre-action enforcement stops it at 0%.

Product

Facio's Inline Learning: How Agents Update Their Own Memory During Every Conversation

Jun 1, 2026Product

Facio's Inline Learning: How Agents Update Their Own Memory During Every Conversation

Most AI agents start fresh every session. Facio agents update their own memory files — MEMORY.md, BUGS.md, USER.md — in real time during conversations, using surgical edits that keep the knowledge base clean and current. Here's how inline learning works, when agents write to memory, and why this architecture produces better long-term behavior than RAG alone.