---

name: lamina-user-modeling description: "Create evidence-grounded Persona and Actor resources and compile independent graph Missions without treating simulations as research."

User modeling

Business documents and .lamina/personas.json are indexable evidence. Canonical Personas and Actors are Resource records in Ladybug:

• Actor expresses authority, ownership, identity proof, and allowed Operations.
• Persona expresses goals, constraints, context, and evidence.
• Link them with a typed Statement such as lamina:canAssume.

Use explicit input, research, analytics, support evidence, or brownfield observations. Keep agent inference epistemically inferred and Persona interpretation simulated. Do not invent demographics or preferences that do not change behavior.

Before implementation, run one independent lamina design prepare-walk task per active Persona and record the isolated result with lamina design record-walk. These design-time simulations walk every proposed operation, including denied and inapplicable nodes, and may expand the graph with missing permissions, states, Scenarios, Invariants, and recovery paths. Rerun all affected walks until a current full round returns empty discovery arrays.

After implementation, compile one independent runtime Mission for every active Persona with lamina mission compile. There is no maximum. Each Mission gets an isolated Run and adapter context. Design simulations and runtime Missions are different evidence classes. Preference findings remain hypotheses; only structural, safety, contradiction, observed, human, or runtime evidence may support stronger claims.