Agateon: verify AI agents the way a build system verifies a compiler

Description and use

Agateon enforces objective gates between AI agent phases, persisting state in Markdown and ensuring that "done" means a reproducible result.

Problem

Long AI agent tasks provide only a vague "it looks done" signal, lacking objective verification.

Outcome

Agateon enforces objective gates between phases, stores state in Markdown, and prevents progress without passing evidence, improving reliability of AI‑driven software tasks.

Workflow

• Define task brief (P0)
• Generate requirements (P1)
• Create design (P2)
• Design tests first (P3)
• Implement code (P4)
• Verify via tests (P5)
• Accept results (P6)
• Ensure consistency (P7)
• Release (P8)
• Mark DONE

Evidence

Agateon is an open‑source orchestration protocol for software‑engineering tasks run by AI agents. There’s no runtime, no daemon, no build step: it’s a set of Markdown protocol files plus gate‑check scripts.

Source

If a gate fails, the phase is redone and the retry is recorded. This is where a recent postmortem — our AI safety net depended on the agent being honest. It wasn’t.

Source