Concept Documentation Index

This page is the entry point for durable SkillPilot concepts. Keep implementation plans and migration workboards in docs/dev/; keep CI rules and review processes in docs/qa-ci/.

Foundations

• Glossary Precise definitions of the core SkillPilot terms, with German equivalents and the normative source per term.
• Levels of Personalization Product-level personalization model.

Skill Graph

• SkillPilot Skill Graph Specification Core graph model and formal vocabulary.
• General Goal System and Migration Long-term goal-system split and migration direction.
• Node Types Node semantics and classification.
• View Projection and Goal Placement Scope-specific learner-facing views and placement semantics.
• Source And Resource Links Provenance and source/reference semantics.
• Atomic Goal Visualizations Production pipeline and review rules for optional atomic-goal image assets.
• Human-Readable Source Rationales Generated explanation model for tracing learning goals back to source evidence.
• Learning-Goal Book and Evidence-Review Pipeline Versioned one-goal-per-page HTML/PDF publication and fast multi-AI evidence-review model with a separate slow public feedback channel.
• Canonical Gymnasium Rollout Policy Conceptual rollout rules for canonical German Gymnasium subjects.
• MEM/FWU Roundtrip Plan Concept plan for MEM/FWU graph interchange.
• Dual Curriculum Package Releases Target architecture for equivalent JSON-runtime and FWU-ontology release packages, standalone consumption, and repository decoupling.

Didactic Concepts

• Feynman Technique for Learning
• Frontier Goal Selection Strategies & Autopilot
• Learning Velocity

Runtime Workflows

• Learning Workflow
• Import/Export Workflow
• Client-State Sync
• Verhaltensintegration des MCP-Lerncoaches Normativer Nordstern, allgemeines Coach-Verhaltensmodell, Golden Journeys und objektive End-to-End-Acceptance.
• Kommunikationsvertrag zwischen ChatClient und SkillPilot-Backend Kanonische, einfache Verantwortungsgrenze für Sprache, Zustand, Tooldesign, Capabilities, Ergebnisse, Fortsetzung und Fehler aller Coach-Adapter.
• SkillPilot-Lerncoach: OpenAI-Plugin-, Skill- und MCP-App-Architektur Target architecture for one multilingual public plugin per contract major, combining a neutral-English coaching skill with a directly submitted MCP server while preserving backend-authoritative interaction language, the SkillPilot domain core, direct provider billing, and staged production gates.
• OpenAI-Plugin: Versionierung und Lebenszyklus Binding version-line, compatibility, snapshot, lifecycle, release, and retirement rules for the multilingual SkillPilot Coach v1 plugin.
• OpenAI-MCP-App: OAuth-Appbindung und 24h-Lernsession Normative separation of the fixed confidential OAuth client that authenticates the App and the fresh, automatically transported, absolute 24-hour learning session that addresses the learner.
• OpenAI-MCP-Clientbindung Security source of truth for client_secret_basic, exact callback/resource/scope binding, secret lifecycle, and the independent learner-session boundary.
• ChatGPT Visible Session Flow Retained Custom-GPT rollback runtime with visible cross-turn relay; not the current production reference architecture.
• Legacy ChatGPT Startcode / Session Flow Historical retained rollback architecture; not the current ChatGPT path.

Maintenance

• Add durable semantic or product concepts to the closest section above.
• Move implementation-only notes to docs/dev/ and operational QA rules to docs/qa-ci/.
• cd app && npm run check:docs-indexes fails if a Markdown file under docs/concept/ is missing from this index.