Skip to content

SkillPilot Quality Assurance: The Four-Level Champion Model

Generated artifact. Do not edit manually.

Generated by: scripts/build_compendia.py Regenerate with: python3 scripts/build_compendia.py --compendium qa-ci --web-and-tex Source of truth: docs/qa-ci/source/

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

Compendium: §1 Maturity & Trials · §2 Scopes & CQR Rules · §3 Graph Validation · §4 Semantic Atomicity · §5 Champion Guide · §6 Four-Level Model

6. SkillPilot Quality Assurance: The Four-Level Champion Model

Forward-looking expansion roadmap

Target architecture for scaling curriculum QA across algorithmic CI, AI agents, and human practice reviews. Currently shipped public maturity and trial badges are governed strictly by 1 (Level 1a and Level 2 are active anchors today, whereas Level 1b and Level 1c represent planned expansion stages).

6.1. Positioning and the Four QA Levels

To scale from AI-derived raw graphs to classroom-ready skill landscapes without burdening domain educators with JSON schemas or DAG invariants, this roadmap unfolds the former Level 1 into three automated/agentic stages (1a, 1b, 1c) ahead of the human practice certificate (Level 2):

Level Current status Inspection focus and representative checks
Level 1a: CI pipeline Active today Content-agnostic algorithmic integrity: acyclicity of contains/requires, schema and ID checks, valid projected learner views (3).
Level 1b: Static AI agent Forward-looking Static didactic and formal guardrails without runtime simulation: route coverage from motivation to autonomy, clean separation of sourceRef and didactic assets, formal completeness of exam and rubric schemas.
Level 1c: Dynamic AI agent Forward-looking Simulated learner walkthroughs under live runtime rules: frontier and filter behavior, dead ends or hidden blockers, exam simulations with incomplete/wrong answers, and SRS Memorize mechanics.
Level 2: Human Champion Active today Authentic pedagogical evaluation in the WebGUI or AI coach: tone, age fit, motivation, subject alignment, and UX — filing issues/PRs toward Human-tested (5).

6.2. Human-Agent Workflow and Governance

The cascade operates as structured human-agent teamwork:

  1. Level 1a blocks mathematically or structurally broken graphs automatically in CI.
  2. Level 1b and Level 1c are designed to pre-triage static and simulated findings at scale.
  3. Level 2 evaluates real pedagogical quality in authentic learning contexts.
  4. Findings are tracked and resolved transparently via GitHub issues and pull requests.
  5. Human maintainers arbitrate edge cases and safeguard pedagogical intent.