Before access
- Finish line and exclusions agreed
- Inputs and data classes named
- Approval and stop conditions written
- Test environment chosen
The artifact is the meeting
The process is written so the buyer can see when scope is fixed, when access expands, how failure is tested, and what survives after delivery.
Async by default. One named buyer. One written review round unless the scope says otherwise.
You should know the finish line before I touch the work. We name the file, screen, decision, or working path that must exist at handoff.
I inspect the actual sources and permissions before asking for more access. That is where constraints and failure cases usually surface.
I build the smallest version that answers the question or makes the path work. Protocols and polish wait until they earn their place.
Then I try the obvious breakpoints: bad input, a rejection, an interruption, and a repeat request. The agreed checks decide whether it is ready.
You get the artifact, its source, the checks I ran, the limits I found, and a manual someone else can follow.
Access expands with proof
Public sources and sandbox fixtures come first. Sensitive production access waits for a written boundary, payment, least privilege, and a named owner.
Scope changes
A new system, screen, decision, access class, or review round becomes a written fixed-price change. Work pauses when required buyer inputs are late.