S15 · Production content tooling, AI and publication
Implemented components; full stream acceptance is not yet demonstrated.
Current implementation: Typed foundations, constrained persistent leads, a real 28-role worker runtime and public portals are implemented.
Outstanding acceptance: This does not claim an autonomous live LLM director or a complete production authoring/admin platform.
The four requirements below preserve the original release target. Their historic baseline is not used as a current completion score. This page does not mark the entire stream complete merely because related components are installed.
Reference
| Requirement | Required deliverable | Acceptance standard |
|---|---|---|
| S15.01 · Build production registry and content authoring | Integrate entities, capability contracts, asset bindings, tests and implementation/deployment evidence. | The admin portal differentiates authored, compiled, deployed and active content; it cannot mark a game feature complete from prose alone. |
| S15.02 · Deploy constrained narrative direction | Implement authorized context, eligible personal opportunities, factual memory, tested fallback text and model/prompt evaluation. | A dialogue model cannot mint items, change mechanics, reveal private answers or stall ordinary gameplay during a model outage. |
| S15.03 · Operate the scoped agent workflow | Configure permissions, isolated branches, owners, review gates, retries and release manifests for the established role contracts. | No agent approves its own economic change or writes production state; outdated review evidence becomes stale. |
| S15.04 · Publish wiki and patch notes from activation | Generate precise mechanics references and reviewed explanations, then reconcile live configuration and public versions. | The public site has no author-only file roots or unreleased acquisition secrets; patch notes describe activated changes, not planned ones. |