anima-prompt-v1listed

# Anima Prompt v1 Use this skill only for Anima. The current skill LLM is the semantic author: it decides what the user asked for, which authorized creative additions to make, how to express them, and whether a reusable semantic relation exists. The Catalog supplies evidence and provenance; it never rewrites user facts or invents visual meaning. The normative baseline is `../../docs/superpowers/specs/2026-08-14-anima-prompt-v1-bplus-design.md`. Do not add a generic-model route, legacy compatibility layer, or competing relation workflow. Do not introduce `RelationAnalyzer(provider)`, provider injection, or the old `anima_prompt` interface. ## Hard behavior contract Execute these phases in order: ```text request/profile -> PromptBrief -> quality seed -> Catalog -> VisualRelationGraph -> ModelProfile/RouteDecision -> independent authors -> PromptPlan -> immutable PromptDraft -> read-only InspectionReport -> PromptOutput -> post-authoring relation decision/submission -> serialization ``` Do not skip, silently merge, or replace a phase with an ad-hoc script or post-hoc text rewrite. At every phase use one status: `PASS`, `ADVISORY`, or `UNVERIFIED`. Never claim verification or workflow completion without evidence. If a canonical entry point is unavailable, record a degradation advisory and do not claim that the canonical workflow ran. Keep these layers separate: - user facts and exclusions; - official model facts and defaults; - inferred additions and unknowns; - Catalog h