Commissioning

Commissioning is the deliberate act of bringing a Post into production. It is not the same as deploying code, turning on an API, or sharing a prompt. Commissioning means The Groundwork is complete, the owning Station is assigned, the Relay triggers are tested, and the agent is formally entered into the operating model.

The word is borrowed from naval and engineering tradition. You commission a vessel — you do not simply push it into the water. The ceremony is the point. It forces the organisation to declare: this system is now infrastructure. It has an owner. It has documents. It has a governance envelope. From this moment forward, changes to its behavior require authorization and testing.

The alternative — deploying an agent and iterating — is the norm right now and it is how organisations accumulate ungoverned automation. Each agent deployed without commissioning is another system running without documentation, without an owner, and without a testing protocol. The technical debt accumulates silently until the first major failure reveals how much of the operation depends on agents nobody formally owns.

Before commissioning, write the Post Brief. One page. What is this Post called? What Station owns it? What is the one-sentence Directive? What is in scope and what is explicitly out? When does it Relay and to whom? How do you know it is working? The brief surfaces questions you have not answered yet, when the cost of answering them is still low.

Then build The Groundwork. scope.md, contracts.md, testing.md, prompting.md. The Groundwork Compiler at waypost.run generates all four in about ten minutes. That output is sixty percent complete and requires another hour or two of specific work.

Then run the gates. Does scope.md clearly define in and out? Does contracts.md freeze the output interface? Does testing.md have golden vectors for edge cases, not just the easy cases? Does prompting.md embed the institutional context the Post needs?

When all four documents pass their gates and the Station Owner signs off, the Post is commissioned. It enters Standard State. It goes on the Post Roster. The governance begins.

The full Groundwork takes real work. For a complex agent in a consequential domain, you are looking at a few hours before the Post runs. That is the standard. It is also not always where you start.

Provisional Commissioning lets you launch a Post with only scope.md and a Directive. The Post runs, but it runs in Pressure State permanently until the remaining documents are completed. Every run generates an entry in the Debt Ledger. Every output requires human review before it ships. No gate gets removed until the full Groundwork is in place.

Provisional Commissioning is not a shortcut that leaves you ungoverned. It is a governed shortcut. The governance is tighter than Standard State because the Post has not earned its autonomy. The system tracks the incomplete documentation, and the human review requirement ensures nothing gets through an unverified gate.

For teams on Day 1 who want to start deploying agents before they have the full protocol dialed in, Provisional Commissioning is the right answer. Get the agent running, get real operational data, build the remaining Groundwork against actual usage rather than hypothetical usage. The Debt builds. You pay it down as you build the documents.