Fair point.. If a website's manifest is wrong any agent trying to use it will just get an error. The agent will learn not to trust that site and the site owner just ends up making their own features unusable. The real incentive is to be useful. Websites that provide an honest manifest are the ones that will actually work with agents.
If someone s goal is simply to break the interaction with agents they can certainly do that by publishing a broken manifest. But that s no different from how the web works today. There will always be a divide between those who build useful services and those who act in bad faith. The solution isnt in the protocol itself, but in the community that uses it. Just as we now have services that warn us about malicious websites a healthy AURA ecosystem would naturally lead to open, community run reputation services. An agent could check a site s aura.json against such a service before trusting it. If a manifest is found to be intentionally misleading its reputation would drop, and agents would collectively learn to ignore it. The system corrects itself through shared information. Trust has to be earned, not just declared in a file.
Fair point.. If a website's manifest is wrong any agent trying to use it will just get an error. The agent will learn not to trust that site and the site owner just ends up making their own features unusable. The real incentive is to be useful. Websites that provide an honest manifest are the ones that will actually work with agents.
…that’s the point
If someone s goal is simply to break the interaction with agents they can certainly do that by publishing a broken manifest. But that s no different from how the web works today. There will always be a divide between those who build useful services and those who act in bad faith. The solution isnt in the protocol itself, but in the community that uses it. Just as we now have services that warn us about malicious websites a healthy AURA ecosystem would naturally lead to open, community run reputation services. An agent could check a site s aura.json against such a service before trusting it. If a manifest is found to be intentionally misleading its reputation would drop, and agents would collectively learn to ignore it. The system corrects itself through shared information. Trust has to be earned, not just declared in a file.