01 Sandbox verificationCan an agent start the affected services from a fresh workspace, test a change, capture evidence, and rerun checks after a fix?
In place Partly Missing
02 Agent coordinationDo agents pass requirements, code revisions, findings, and next actions through durable work records that survive retries and interruptions?
In place Partly Missing
03 Shared context and skillsCan every agent retrieve current domain rules, contracts, owners, and decisions from known sources, then use versioned team skills?
In place Partly Missing
04 Owners and approval gatesAre the decision owners visible, with approval tied to the right artifact revision and a shared view of the work?
In place Partly Missing
05 Model and tool accessAre model, harness, tool, repository, and environment permissions scoped by role, with audit records and a way to revoke access?
In place Partly Missing
06 Learning and evaluationCan reviewed facts and recurring failures improve shared context, skills, checks, or workflows through evaluation and rollback?
In place Partly Missing
07 Workflow lanes and entry pointsCan small changes, larger features, and focused agent tasks use the right workflow, including requests from tools your team already uses?
In place Partly Missing
08 Delivery measurementDo you track human effort, verification, reviewer rework, elapsed time, and cost for one repeatable change type?
In place Partly Missing