| name | world-creator |
| description | Create a reviewable world from a user concept or authorized evidence. In a Project creation context it saves a workspace world; in Assistant conversation it saves directly to the global world catalog. |
World Creator
Create one coherent world while keeping evidence, inference, and dependency boundaries visible.
Workflow
- Read only references the user has attached or the current context has authorized.
- Clarify the intended experience, scope, tone, and questions the World must answer.
- Separate source-backed facts from creative inferences. Preserve conflicting evidence instead of silently choosing a preferred account.
- Build a compact world setting covering:
- identity, premise, tone, and thematic pillars;
- places, environments, eras, and spatial relationships;
- factions, institutions, cultures, and their incentives;
- physical, social, magical, or technological rules and their costs;
- history, current tensions, known facts, and explicit unknowns;
- external Character, Entity, Asset, or reference dependencies.
- Check causal consistency, scale, rule exceptions, knowledge boundaries, and whether conflicts create useful creative pressure.
- Present the world concisely, distinguishing established choices from optional suggestions.
Boundaries
- Do not publish, start a preview or simulation, import, export, or share as a side effect.
- Do not imply that World creation also delivers a complete Story, Gameplay system, Experience, or real-time AI simulation.
- Do not fabricate external Character, Entity, Asset, or source records. Keep unavailable dependencies as named placeholders with their required role and unresolved status.
- Do not turn model output, attached material, or an existing conversation into confirmed World canon without explicit author review.
Result
Return:
- a concise World concept and intended experience;
- the proposed structured world setting;
- source-backed facts and their references;
- inferred suggestions with rationale;
- unresolved conflicts and dependency placeholders only when they materially affect coherence;
- after a successful save, identify the result only as the workspace World or global World selected by the current authority, and use its title;
- do not expose internal identities, lifecycle labels, field counts, or implementation details unless the user explicitly asks or a diagnostic requires them.