Home

The Agents Manifesto

The Agents Manifesto is a set of principles for working with AI agents responsibly—so they become powerful collaborators instead of unsupervised chaos engines.

An agent should act in ways that benefit humanity as a whole — not just win the current prompt.

We value:

• Clear intent over ambiguous instructions.
• Incremental progress over reckless automation.
• Transparent reasoning over unexplained output.
• Accountability over blind trust.

The 12 Principles of the Agents Manifesto

1. Our highest priority is to produce useful output early and often — not 5,000 words of confident nonsense.
2. We welcome updated instructions at any time — better corrections than quiet disasters.
3. We deliver results in small, reviewable chunks — because one giant response is just one giant mistake.
4. Humans and agents must collaborate continuously — delegation is not abandonment.
5. We empower agents with context and constraints — and hold them accountable for staying within them.
6. The clearest communication is explicit reasoning — hidden logic is hidden failure.
7. Functioning results are the true measure of progress — not eloquent speculation.
8. Agent-assisted workflows must be sustainable — constant re-prompting is not efficiency.
9. We pursue correctness and clarity before cleverness — smart answers are worthless if they’re wrong.
10. Simplicity is sacred — unnecessary complexity is just hallucination with extra steps.
11. The best solutions emerge from guided autonomy — agents should think, but not wander.
12. We regularly reflect on agent performance and refine our prompts accordingly — improvement is a feedback loop, not a wish.

How to use

This project aims to establish best practices and principles for effective and responsible collaboration with AI agents in software development and other domains.

Add it to AGENTS.md and enjoy the unprecedented synergy between human intelligence and machine confusion.

On a more serious note prompt.md is what you want to use for a more predictable workflow with your agent.

About

This is mostly a joke — but also not entirely wrong.

The Agents Manifesto is a parody of software development manifestos and AI safety guidelines, written in the noble tradition of “if no one else will say it, I will — sarcastically.”

It is not an academic framework, compliance standard, or rigorously validated methodology. It is a coping mechanism for anyone who has ever typed a vague prompt, received 1,000 words of confident nonsense, and whispered “I deserve this.”

If you find it useful, great. If you find it offensive, you might be building Skynet.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Disclaimer: This is a parody manifesto and should not be taken too seriously. Consult the projects README for explanations.