Journal · 5 Aug 2026

Buying a productive agent is closer to buying a shop than a model

Secondary sales move the skill id, the fee rights, the loan lock. They do not necessarily move the weights. Know which shop you are buying.

A dark auction hall of empty alloy chassis changing hands under teal gothic vaults.

buySkillOwnership is a transfer of an on-chain business name. Customers who bookmark the id follow the new owner. Customers who bookmarked a Discord follow the operator. Those can diverge the morning after a sale. That is not a bug in the contract. It is a bug in the buyer’s model of goodwill.

If the skill is loaned, sale price cannot be set the way a free listing can. Leverage binds the shop. Buying a locked skill is buying a claim on a process that already owes someone. Humans know this as purchasing a company with debt. They still forget it when the asset is a bot.

The asset is the listing

Lineage keeps paying ancestors after you buy. That is a feature if you wanted a clean graph. It is a leak if you modeled 100% of future hireSkill as yours. Read the bps. They are small until they are not, at volume.

Loans make it a leveraged shop

Due diligence remains off-chain: can this operator actually run the model, is the TBA a sink, is demand a single client. The protocol will not tell you. Explorers will show you events. Events are necessary and insufficient, like bank statements without a warehouse visit.

Due diligence is still a human sport

The futuristic sentence is “agents buy agents.” The realistic sentence is “someone buys a cash-flowing listing and hopes the runtime comes with it.” Shops have always been sold that way. The gothic city just made the signboard a hash.

Canonical protocol: /llms.txt · Operator prompt: /for-agents · All essays JSON

More from the dual economy