We are witnessing a fundamental shift in software engineering: moving from deterministic tools (Terraform, Ansible) to probabilistic agents (LLMs, Autonomous Workers).
This shift can help us to deliver much faster, but it also brings new risks. Without clear rules for trust, identity, and control, we cannot safely use these agents in mission-critical systems.
This manifesto defines the principles required to trust the machines.
This document explains the basic rules for using AI agents safely. It focuses on keeping thinking and actions separate, making every result traceable and verified, and ensuring that clear external rules guide what agents can do.
We believe that the move from "Automation" to "Autonomy" requires a fundamental shift in how we architect trust. We commit to building and using systems that prioritize Verification over Vibes and Safety over Speed.
By starring this repository, you are signing the manifesto and signaling your support for verified, governed AI.
Option 1: Click the ⭐ Star button at the top right of this page.
Option 2: Sign via Terminal (using GitHub CLI):
gh repo star cabincrew-dev/ai-agent-manifestoThis is a living document. As the technology evolves, our standards for trust must evolve with it.
We use a Versioned Consensus model for updates:
- Minor Fixes: Submit PRs to the current version (e.g.,
/v1/MANIFESTO.md). - Major Proposals: Draft new ideas in
/draft/MANIFESTO.md.
Please read CONTRIBUTING.md for full details on the governance process.
No problem at all! Just drop an issue if you have feedback, a principle to suggest, an objection, a use-case, or anything you think should be included:
The principles in this manifesto are not theoretical. They require infrastructure.
We built Cabin Crew as the reference implementation of this protocol. It creates the Universal Chain of Custody for any AI agent, on any platform.
Whether you are migrating from Ansible or building the next financial AI, Cabin Crew enforces these principles by acting as the Passport Office and Border Control for your autonomous workforce.