Skip to content

Atomic Goal Visualizations

Generated artifact. Do not edit manually.

Generated by: scripts/build_compendia.py Regenerate with: python3 scripts/build_compendia.py --compendium skill-graph-architecture --web-and-tex Source of truth: docs/concept/skill-graph/architecture-source/

Language / Sprache: English (EN) · Deutsch (DE) | PDF: English PDF · Deutsches PDF

Compendium: §1 General Goal System · §2 Node Types · §3 View Projection · §4 Source & Resource Links · §5 Content Integration · §6 Source Rationales · §7 Goal Visualizations · §8 Learning-Goal Book · §9 Gymnasium Rollout · §10 MEM/FWU Plan · §11 Dual Package Releases

7. Atomic Goal Visualizations

Optional concept illustrations help learners grasp the core idea of an atomic learning goal visually without replacing independent reasoning or task solving.

Reviewed visualizations attach to canonical atomic goals via resourceLinks:

{
  "type": "goal-visualization",
  "resourceType": "image",
  "role": "primary",
  "skillpilotId": "<exact-canonical-goal-id>",
  "url": "/assets/goal-visualizations/<subject>/<id>/<id>.png"
}

In MCP-capable hosts such as OpenAI ChatGPT, render_skillpilot_goal_visualization serves a surface-neutral text/html;profile=mcp-app widget governed by a one-shot (goalId, stateVersion) presentation authorization and the learner toggle showGoalVisualizationsInChat. The widget remains collapsed until the image loads cleanly, accepts bounded MCP result envelopes while using persisted widget state only as a bootstrap fallback, and retains all previously advertised content-addressed UI resource hashes unchanged.

7.2. Production Workspace, Format, and Provider Policy

Authoring prompts, candidate generation, and QA ledgers reside under curricula/DE/Gymnasium/visualizations/ (prompt.de.md, image-reconstruction-prompt.de.md, published via npm run deploy:assets at 16:9 1600x900 PNG and inspected at both 360px mobile and 680px desktop widths, calibrated on pilot goal 502ecaa7-cca6-5c51-a1cc-da09a7b2382c).

Under the binding Product Owner decision of 13 September 2026:

  • Default KEEP: Retain good existing Mathematics and Physics images regardless of which provider generated them; never regenerate retro-actively merely to unify providers.
  • Provider priority for new or defective images: Prefer ChatGPT/Codex image generation in SkillPilot's friendly, abstract, clear, comic-like raster style when subject-matter accuracy and representation fit. Nano Banana Pro (gemini-3-pro-image) is an allowed alternative whenever it better serves the didactic representation — no prior failed attempt with the other provider is required.
  • Narrow SVG/programmatic exception: Programmatically plotted diagrams or SVGs remain narrow, explicitly documented exceptions justified by exact geometric/function-plot precision, never a blanket style replacement.

7.3. CLI Workflow and Two-Stage Review Gates

Helper scripts in app/ (visualization:generate:nano-banana, visualization:plan-batch, visualization:check-prompt-appends, visualization:prepare, visualization:import) separate asset generation from approval:

  1. AI QA ledger (aiApprovedAssetSha256 / Gate V): Records structured review labels (accepted_pilot, accepted_pilot_after_regeneration, rejected_regenerate, deferred_provider_limitation, needs_external_review) bound to exact file SHA-256 digests.
  2. Human publication gate (humanApproved, HR-003): Public Standalone releases and public-mode GoalBooks (8, Human Review Gates) require explicit human content and rights clearance (CC-BY-4.0 vs. provenance); generation or AI review alone never grants publication approval.