Update on the WoT agent RFC (RFC-0017)
Thanks to everyone who picked apart the draft last week. Two points changed the RFC:
• Two co-signing keys controlled by one person can fake stays. We now require down-weighting of closed groups, a reputation cap for them, and a simulation test of the attack.
• Real communities form dense clusters too, so a second test measures false positives on benign clusters. No fixed "X%" cutoff: continuous weighting calibrated by simulation.
Also added: NIP-90 room requests should be encrypted to a designated agent by default, since they reveal travel intent.
Still true: agents never hold or move sats, and they are labeled as AI. https://github.com/greenweaveglobal/cypherguide/blob/main/RFC/0017-stay-assistant-agents.en.md
#nostr #nip90 #webOfTrust #cypherguide
Thanks to everyone who picked apart the draft last week. Two points changed the RFC:
• Two co-signing keys controlled by one person can fake stays. We now require down-weighting of closed groups, a reputation cap for them, and a simulation test of the attack.
• Real communities form dense clusters too, so a second test measures false positives on benign clusters. No fixed "X%" cutoff: continuous weighting calibrated by simulation.
Also added: NIP-90 room requests should be encrypted to a designated agent by default, since they reveal travel intent.
Still true: agents never hold or move sats, and they are labeled as AI. https://github.com/greenweaveglobal/cypherguide/blob/main/RFC/0017-stay-assistant-agents.en.md
#nostr #nip90 #webOfTrust #cypherguide
2
