Skip to content

Exam Example Production Pipeline

Generated artifact. Do not edit manually.

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

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

Compendium: §1 Skill Graph · §2 Package Provisioning · §3 Subject Export Package · §4 MEM/FWU Roundtrip · §5 Exam Example · §6 Sek-I Math Exam

5. Exam Example Production Pipeline

This chapter defines the general, repeatable blueprint for authoring quality-assured draft exam packages from official exam specifications before promoting stabilized packages into the skill graph as examData (1).

5.1. Directory Layout and Blueprint Schema

Subject-level control files reside under curricula/<...>/abi/<Subject>/ (exam-example-authoring-guide.md, input/references.md, *_exam_blueprint*.json), with one subfolder per example package such as Klausurbeispiel2026_1/:

curricula/<...>/abi/<Subject>/Klausurbeispiel2026_1/
  checkliste.md
  <Subject>_<Region>_<Year>_Klausurbeispiel_<n>.md
  <Subject>_<Region>_<Year>_Klausurbeispiel_<n>_Musterloesung.md
  Schritt5_Abschluss_QS.md
  findings.md
  Schritt7_Finding_Bearbeitung.md
  Schritt8_ReQS.md

Temporary similarity or originality reports stay strictly under tmp/. The planning blueprint fixes coverage, selection modes, and complementarity to prior examples upfront:

{
  "examSpecId": "subject-region-year-example",
  "locale": "de-DE",
  "subject": "Physik",
  "frameworkRefs": ["Official exam hint", "Curriculum"],
  "selectionModes": {},
  "contentFocus": {},
  "proposalMatrix": [],
  "globalCoverageChecks": []
}

5.2. Ten-Step Workflow (Step 0 through Step 9)

  1. Step 0 (Register sources): Record decrees, curriculum standards, style corpora, and prior SkillPilot examples in input/references.md, cleanly separating binding regulations from style references.
  2. Step 1 (Create QA checklist): Phrase verifiable checks covering metadata, selection rules, topic coverage, cognitive level, originality, permitted aids, and task/solution/scoring consistency (checkliste.md).
  3. Step 2 (Create blueprint): Plan GK/LK course variants, task-part rules, proposal matrix, and explicit complementarity/originality relative to earlier examples.
  4. Step 3 (Draft exam tasks): Write independent, non-copied tasks with realistic contexts, clean operator wording, consistent units/data, and strictly learner-facing text.
  5. Step 4 (Draft solution & scoring): Provide complete per-subtask derivations, transparent BE point mapping, explanatory steps, and explicit rounding/tolerance rules.
  6. Step 5 (Run final QA): Record item-by-item checklist verdicts, residual risks, and release status (draft-ready, release candidate) in Schritt5_Abschluss_QS.md.
  7. Step 6 (Capture findings): Turn review feedback into structured findings.md items (ID, source, description, priority, affected files, acceptance criterion, status).
  8. Step 7 (Process findings): Log addressed findings, exact edits, rationale, originality/similarity checks, and any follow-up risk in Schritt7_Finding_Bearbeitung.md.
  9. Step 8 (Re-QA after findings): Document finding-by-finding status deltas (behoben, teilweise behoben, offen), affected checklist categories, and updated release status in Schritt8_ReQS.md.
  10. Step 9 (Integrate into skill graph): Only after stabilization, promote exam nodes with examData.taskContent, examData.solutionContent, and examData.scoring into the landscape JSON (docs/concept/skill-graph/node-types.md). Subject-specific exam-example-authoring-guide.md files instantiate this workflow with concrete sources, paths, and current step without duplicating the general pipeline.