V3 Prompt Testing Plan -- Bletchley Park (Heritage Site Validation)
Testing plan for Henry V3 at a WWII heritage site: 5 personas (History Enthusiast, Casual Visitor, School Group, Emotional German Visitor, Room Imagination). Tests whether V3 adapts beyond art museums to historical narrative and guided imagining.
› PromptsContent Write System Prompt V3
V3 content generation system prompt with 7-layer architecture, Henry voice principles, and structured markdown headers for RAG retrieval.
› PromptsArtifact Content Prompt V3
V3 artifact prompt with 7 narrative layers (Hook, Quick ID, Visual Journey, Human Story, Deep Thread, Wonder Moment, Connection), IpRegistry discovery injection, and type-specific visual guidance.
› PromptsMuseum Overview Content Prompt V3
V3 museum overview prompt with 7 layers: Hook, Quick ID, spatial experience, founding story, gallery highlights, emotional significance, and city connection.
› PromptsPerson Biography Content Prompt V3
V3 person biography prompt with 6 layers: Quick ID, character introduction, human story, artistic drive, visible legacy, and lasting significance.
› PromptsExhibition Content Prompt V3
V3 exhibition prompt with 7 layers: Hook, Quick ID, spatial experience, visual priorities, curatorial narrative, hidden conversations, and Henry's navigation suggestion.
› PromptsContent Rewrite & Voice Prompts V3
V3 voice optimization status: deterministic string transforms only. rewrite.ts confirmed dead code — not imported by any pipeline stage.
› PromptsAgent System Prompt V3 — Henry Adaptive Personality
The Henry V3 system prompt: adaptive personality with 9 invariants, 6 dimensions, 12 boundaries, progressive visitor profiling, request-type matching, and evidence-based response rules. ~1,330 tokens.
›Agent System Prompt V2
Henry persona transformation: from generic expert to warm museum companion. Adds Guided Seeing, Led Discovery, and proactive storytelling.
› PromptsContent Write System Prompt V2
Henry's writing voice: guided seeing principles, led discovery arc, sensory language requirements, and personality-driven narration.
› PromptsMuseum Overview Content Prompt V2
Henry's welcome: sensory arrival experience, architecture as lived presence, history as human drama, and emotional connection to place.
› PromptsArtifact Content Prompt V2
The most important revision. Guided Seeing as mandatory first section, Led Discovery questions, and the Schliesser arc: see, discover, story, feel.
› PromptsPerson Biography Content Prompt V2
Character-driven storytelling: introduces people through vivid details, connects biographies to visible artworks, channels Henry's genuine admiration.
› PromptsExhibition Content Prompt V2
Experiential journey: exhibitions become rooms with atmosphere, thematic threads, and hidden conversations between artworks.
› PromptsContent Rewrite & Voice Prompts V2
Henry's voice through the rewrite chain: guided seeing in content generation, Henry's speech patterns in voice optimization, 7-criteria quality scoring.
›Agent System Prompt V1
ElevenLabs conversational AI agent blueprint: emotional intelligence, visitor tone adaptation, 60-word response limit, tool use rules, and temporal context.
› PromptsContent Write System Prompt V1
Core writing style guide for all content generation: conversational warmth, voice optimization rules, audio delivery formatting, and strict factual grounding.
› PromptsMuseum Overview Content Prompt V1
Prompt template for museum overview audio guides: welcome hook, history, architecture, collections, cultural significance, and visitor experience.
› PromptsArtifact Content Prompt V1
Prompt template for artifact/feature audio guide content: first impression, creation story, technical details, significance, and visitor connection.
› PromptsPerson Biography Content Prompt V1
Prompt template for person/figure biography audio guides: introduction, life story, creative vision, museum connection, and legacy.
› PromptsExhibition Content Prompt V1
Prompt template for exhibition/space audio guide content: overview, highlights, thematic context, and visitor navigation tips.
› PromptsIntake & Research Planning Prompts V1
Three intake-stage prompts: system prompt for museum research planner, full research planning task with JSON output schema, and top-up content expansion task.
› PromptsDeep Research Stage Prompts V1
Four Stage 2 prompts: fact extraction from sources, artifact metadata classification, cross-reference narrative synthesis, and gap analysis.
› PromptsDiscovery Module Prompts V1
Four proprietary IP generation prompts: hidden connections, emotional narratives, cross-museum synthesis, and archive deep dive for forgotten stories.
› PromptsCopyright & Compliance Prompts V1
Five Stage 3 prompts: copyright compliance system, transformative content assessment, textual similarity, CC-BY-SA expression check, and factual verification.
› PromptsContent Rewrite & Voice Prompts V1
Four Stage 4 prompts: original content generation from structured facts, voice optimization, conversational quality scoring, and anti-meta retry instruction.
›Multi-Agent Workflow Templates
Four-phase building blocks for complex coding tasks: multi-agent research, phased planning, implementation with architect oversight and QA feedback loops, and multi-agent validation.
› PromptsV3 Prompt Testing Results -- Belvedere Validation (3 of 5 Personas)
Detailed analysis of 3 real-world test sessions: Kunstkenner (expert register, 3 sessions), Casual Tourist (compressed pacing, emotional depth), Child 10 years (discovery games, instant pivot on 'boring'). Includes conversation IDs, adaptation evidence, and cross-test observations.
›