chore(tooling): bring template-owned files in line with the template (#33)

Copies every file .github/ownership.json gives to the template byte for
byte from adaptive-learner-content-template (0b2971a): the dictionary-based
prose gate, quality decisions read from .github/quality-state.json, the
shared adopted ext: list with its nightly comparison against the app, and
the nightly ownership check.

.github/quality-state.json records the prose backlog this repository
carries (non-blocking, with date and reason); the gate's self-test stays
blocking. No lesson, manifest or pin changes. The search index files are
owned by the hub and follow once its generator and the template's are
merged there.

Co-authored-by: Asterios Raptis <asteri.raptis@gmail.com>
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Author: 3 people

.github/ownership.json

@@ -0,0 +1,103 @@
+{
+  "description": "Which side owns each path of a content repository. Read by scripts/check_ownership.py (see its docstring for the owner kinds); the Ownership workflow runs it nightly in every repository.",
+  "owners": {
+    "template": "astrapi69/adaptive-learner-content-template",
+    "hub": "astrapi69/adaptive-learner-content"
+  },
+  "paths": {
+    ".github/ownership.json": "template",
+    ".github/workflows/engine-currency.yml": "template",
+    ".github/workflows/engine-validate.yml": "template",
+    ".github/workflows/ownership.yml": "template",
+    ".github/workflows/registry-drift.yml": "template",
+    ".github/workflows/schema-drift.yml": "template",
+    ".github/workflows/stable-ids.yml": "template",
+    ".github/workflows/umlaut-data.yml": "template",
+    ".github/workflows/validate-content.yml": "template",
+    ".gitignore": "template",
+    "Makefile": "template",
+    "docs/LESSON-FORMAT.md": "template",
+    "docs/ai-review-prompt-template.md": "template",
+    "generated/README.md": "template",
+    "requirements.txt": "template",
+    "schema/README.md": "template",
+    "scripts/adopted-extensions.mjs": "template",
+    "scripts/audit_content.py": "template",
+    "scripts/build_umlaut_stems.py": "template",
+    "scripts/check_adopted_extensions.mjs": "template",
+    "scripts/check_ownership.py": "template",
+    "scripts/check_prose.py": "template",
+    "scripts/check_schema_drift.py": "template",
+    "scripts/export_anki.py": "template",
+    "scripts/export_quiz_pdf.py": "template",
+    "scripts/export_set.py": "template",
+    "scripts/generate_exercises.py": "template",
+    "scripts/lce_schema.py": "template",
+    "scripts/quality_state.py": "template",
+    "scripts/umlaut_stems.json": "template",
+    "scripts/validate_content.py": "template",
+    "scripts/validate_with_engine.mjs": "template",
+    "templates/README.md": "template",
+    "templates/knowledge/lesson.json": "template",
+    "templates/language/lesson.json": "template",
+    "templates/programming/lesson.json": "template",
+    "tests/fixtures/graded-quiz-sample.json": "template",
+    "tests/fixtures/lesson-shape-parity.json": "template",
+    "tests/test_audit_content.py": "template",
+    "tests/test_check_ownership.py": "template",
+    "tests/test_check_prose.py": "template",
+    "tests/test_check_schema_drift.py": "template",
+    "tests/test_export_anki.py": "template",
+    "tests/test_export_quiz_pdf.py": "template",
+    "tests/test_from_cards_pairs_gate.py": "template",
+    "tests/test_generate_exercises.py": "template",
+    "tests/test_hint_length_rule.py": "template",
+    "tests/test_mc_only_exemption.py": "template",
+    "tests/test_quality_state.py": "template",
+    "tests/test_shape_parity.py": "template",
+    ".github/workflows/generate-index.yml": "hub",
+    "schema/search-index.schema.json": "hub",
+    "scripts/generate_search_index.py": "hub",
+    "tests/test_generate_search_index.py": "hub",
+    "tests/test_search_index_ai_validated.py": "hub",
+    "tests/test_search_index_review_status.py": "hub",
+    "tests/test_search_index_schema.py": "hub",
+    "tests/test_search_index_updated_at.py": "hub",
+    "tests/test_search_index_visibility.py": "hub",
+    ".github/workflows/ai-review.yml": "hub-only",
+    ".github/workflows/validate-registered-repos.yml": "hub-only",
+    "CONTRIBUTING.md": "hub-only",
+    "docs/CROSS-REPO-SEARCH.md": "hub-only",
+    "docs/REGISTER-A-REPO.md": "hub-only",
+    "examples/inception-effekt/README.md": "hub-only",
+    "examples/inception-effekt/lesson.json": "hub-only",
+    "recommended-repos.json": "hub-only",
+    "schema/recommended-repos.schema.json": "hub-only",
+    "scripts/ai_review.py": "hub-only",
+    "scripts/validate_registered_repo.py": "hub-only",
+    "scripts/validate_registry.py": "hub-only",
+    "tests/test_ai_review.py": "hub-only",
+    "tests/test_validate_registry.py": "hub-only",
+    ".github/quality-state.json": "repo",
+    "LICENSE": "repo",
+    "README.md": "repo",
+    "books.yaml": "repo",
+    "docs/GETTING-STARTED.md": "repo",
+    "docs/export-set-usage.de.md": "repo",
+    "docs/export-set-usage.md": "repo",
+    "docs/review/*": "repo",
+    "manifest.yaml": "repo",
+    "media.yaml": "repo",
+    "schema/engine-version.txt": "repo",
+    "schema/stable-id-coverage.txt": "repo",
+    "search-index.json": "repo",
+    "sets/*": "repo",
+    "tests/test_export_set.py": "repo",
+    "tests/test_known_domain.py": "repo",
+    "tests/test_prompt_duplication_rule.py": "repo",
+    "tests/test_readme_set_table.py": "repo",
+    "schema/content-manifest.schema.json": "engine",
+    "schema/lesson.schema.json": "engine",
+    "schema/quality-rules.json": "engine"
+  }
+}

.github/quality-state.json

@@ -0,0 +1,7 @@
+{
+  "prose_gate": {
+    "blocking": false,
+    "since": "2026-09-23",
+    "reason": "Rückstand aus der Zeit vor dem Gate: 169 Em-Dashes und 19 Ersatzschreibungen für Umlaute in 33 Dateien, gemessen am 2026-09-23 mit den Wörterbuch-Daten des Templates (adaptive-learner-content-template#75). Der Rückstand liegt in Dateien, die dem Repo gehören (Lektionen, Manifeste, eigene Doku und Tests), und wird im Inhaltsdurchgang dieses Repos abgebaut; danach wird hier blocking auf true gesetzt."
+  }
+}

.github/workflows/engine-validate.yml

@@ -44,6 +44,8 @@ jobs:
         # a newly pinned engine release becomes visible on its own: warnings
         # otherwise reach nobody unless someone runs `make lint-warnings` by
         # hand, so a pin bump changes a version number and nothing on screen.
+        # The findings go into the job summary, readable without opening the
+        # log.
         continue-on-error: true
         run: |
           # Keep the runner's exit code: it is 0 when it only finds warnings, so
@@ -54,14 +56,11 @@ jobs:
           status=$?
           set -e
           cat /tmp/lint-warnings.txt
+          # Per rule, and for every warning the repository decided to keep
+          # (.github/quality-state.json) the decision next to the number and a
+          # comparison of the recorded elements with the current ones.
+          python3 scripts/quality_state.py warnings /tmp/lint-warnings.txt >> "$GITHUB_STEP_SUMMARY"
           {
-            echo "### Author lints (warnings, non-blocking)"
-            echo
-            echo "| rule | count |"
-            echo "|---|---|"
-            grep -oE '\[W-[A-Z0-9-]+\]' /tmp/lint-warnings.txt \
-              | sort | uniq -c | sort -rn \
-              | awk '{gsub(/[][]/, "", $2); print "| `" $2 "` | " $1 " |"}' || true
             echo
             echo "<details><summary>full list</summary>"
             echo

.github/workflows/ownership.yml

@@ -0,0 +1,120 @@
+# Ownership: every copy of an owned file is its owner's bytes.
+#
+# .github/ownership.json names the owner of every path: the template (author
+# tooling), the hub adaptive-learner-content (search index and federation),
+# the hub alone (registry, AI review), the repository itself, or the engine
+# (the schema mirror, compared by schema-drift.yml). scripts/check_ownership.py
+# compares this repository's copies with the owners' main branches, using the
+# template's current ownership file.
+#
+# Like the registry drift, this does not belong in the PR gate: an owner moves
+# in its own repository, so a PR here would turn red for something this repo
+# did not do. It runs nightly, opens or updates one issue while copies differ,
+# and closes it when they agree again.
+name: Ownership
+
+on:
+  schedule:
+    - cron: "15 6 * * *"
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  issues: write
+
+env:
+  FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
+  OWNERSHIP_LABEL: ownership-drift
+
+jobs:
+  ownership:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/checkout@v5
+        with:
+          repository: astrapi69/adaptive-learner-content-template
+          ref: main
+          path: .owners/template
+      - uses: actions/checkout@v5
+        with:
+          repository: astrapi69/adaptive-learner-content
+          ref: main
+          path: .owners/hub
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.12"
+
+      - name: Compare with the owners
+        id: compare
+        run: |
+          set +e
+          python3 .owners/template/scripts/check_ownership.py \
+            --template .owners/template --hub .owners/hub \
+            --repository "$GITHUB_REPOSITORY" --root . > /tmp/ownership-report.txt 2>&1
+          status=$?
+          set -e
+          cat /tmp/ownership-report.txt
+          # 0 = in line, 1 = findings, anything else = the check itself is
+          # broken (owner checkout or ownership file missing or malformed) and
+          # must fail the job loudly.
+          if [ "$status" -gt 1 ]; then exit "$status"; fi
+          echo "drift=$([ "$status" -eq 1 ] && echo true || echo false)" >> "$GITHUB_OUTPUT"
+          {
+            echo "### Ownership"
+            echo
+            echo '```'
+            cat /tmp/ownership-report.txt
+            echo '```'
+          } >> "$GITHUB_STEP_SUMMARY"
+
+      - name: Open or update the ownership issue
+        if: steps.compare.outputs.drift == 'true'
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          gh label create "$OWNERSHIP_LABEL" \
+            --description "A copy of an owned file differs from its owner" \
+            --color ededed 2>/dev/null || true
+          title="chore(tooling): copies differ from their owners"
+          body="$(cat <<EOF
+          \`.github/ownership.json\` (the template's current copy) names the
+          owner of every path. This repository's copies do not match:
+
+          \`\`\`
+          $(cat /tmp/ownership-report.txt)
+          \`\`\`
+
+          - \`differs from template\` / \`differs from hub\`: bring the copy in
+            line with the owner, or, if this repository's version is the
+            better one, move the change into the owner first.
+          - \`missing\`: copy the owner's file here.
+          - \`not in <owner>\`: the owner does not carry a file the ownership
+            file gives it; fix the owner or the row.
+          - \`only the hub carries it\`: remove it here, or give it a
+            different owner.
+          - \`no owner\`: add a row to \`.github/ownership.json\` in the
+            template.
+
+          This issue is maintained by \`.github/workflows/ownership.yml\` and
+          closes itself once the copies agree.
+          EOF
+          )"
+          existing="$(gh issue list --label "$OWNERSHIP_LABEL" --state open --json number --jq '.[0].number')"
+          if [ -n "$existing" ]; then
+            gh issue edit "$existing" --title "$title" --body "$body"
+            echo "updated issue #$existing" >> "$GITHUB_STEP_SUMMARY"
+          else
+            gh issue create --title "$title" --body "$body" --label "$OWNERSHIP_LABEL"
+          fi
+
+      - name: Close the ownership issue once the copies agree
+        if: steps.compare.outputs.drift == 'false'
+        env:
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          existing="$(gh issue list --label "$OWNERSHIP_LABEL" --state open --json number --jq '.[0].number' 2>/dev/null || true)"
+          if [ -n "$existing" ]; then
+            gh issue close "$existing" \
+              --comment "Every owned copy matches its owner again. Closed by the ownership check."
+          fi

.github/workflows/umlaut-data.yml

@@ -0,0 +1,47 @@
+# Umlaut data: the prose gate's umlaut check against the dictionaries.
+#
+# scripts/umlaut_stems.json is generated from the German and English
+# dictionaries (scripts/build_umlaut_stems.py) and committed; the PR gate only
+# reads it. The tests that hold the committed data and the real gate against
+# those dictionaries (every umlaut word found, no correct German or English
+# word flagged) need the dictionary packages, which the regular test step does
+# not install. They run here, only when the gate or its data changes, and with
+# REQUIRE_DICTIONARIES set a missing dictionary fails instead of skipping.
+name: Umlaut data
+
+on:
+  pull_request:
+    paths:
+      - scripts/check_prose.py
+      - scripts/build_umlaut_stems.py
+      - scripts/umlaut_stems.json
+      - tests/test_check_prose.py
+      - .github/workflows/umlaut-data.yml
+  push:
+    branches: [main]
+    paths:
+      - scripts/check_prose.py
+      - scripts/build_umlaut_stems.py
+      - scripts/umlaut_stems.json
+      - tests/test_check_prose.py
+      - .github/workflows/umlaut-data.yml
+
+env:
+  FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
+
+jobs:
+  dictionaries:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
+        with:
+          python-version: "3.12"
+      - name: Install dictionaries
+        run: sudo apt-get update && sudo apt-get install -y wngerman wamerican wbritish
+      - name: Install test dependencies
+        run: pip install "pytest>=8,<9" regex
+      - name: Gate against the dictionaries
+        env:
+          REQUIRE_DICTIONARIES: "1"
+        run: python -m pytest tests/test_check_prose.py -q

.github/workflows/validate-content.yml

@@ -33,36 +33,24 @@ jobs:
       - name: Install validator dependencies
         run: pip install "pyyaml>=6,<7" "jsonschema>=4,<5" "pytest>=8,<9" regex "genanki>=0.13,<0.14"
       - name: Prose gate self-test
-        # Blocking even while the backlog below is open: a gate that does not
-        # fire on known-bad input is broken, backlog or not.
+        # Blocking always: a gate that does not fire on known-bad input is
+        # broken, whatever the repository's backlog.
         run: python scripts/check_prose.py --self-test
-      - name: Prose findings (non-blocking until the cleanup PR)
-        # This repository still carries a backlog the gate reports: em dashes,
-        # invisible characters, German written with letter pairs instead of
-        # its umlauts. The per-repo cleanup PR works it off and removes
-        # continue-on-error in the same commit; from then on a regression
-        # fails the PR. Until then the step shows red and the job stays green.
-        continue-on-error: true
+      - name: Prose gate (em dash, invisible characters, substituted umlauts)
+        # Blocking unless the repository records a backlog in its own
+        # .github/quality-state.json (prose_gate.blocking: false, with a date
+        # and a reason). The decision lives there and not in this file, so
+        # this workflow stays identical in every repository. A broken gate
+        # (any exit code other than 0 or 1) and a malformed state file always
+        # fail. The summary lists the findings per file and ends with the
+        # gate's totals.
         run: |
           set +e
           python -u scripts/check_prose.py > /tmp/prose.txt 2>&1
           status=$?
           set -e
           cat /tmp/prose.txt
-          {
-            echo "### Prose gate (non-blocking until the cleanup PR)"
-            echo
-            echo "| file | findings |"
-            echo "|---|---|"
-            grep -E '^[^ ].*:[0-9]+: ' /tmp/prose.txt | cut -d: -f1 | sort | uniq -c | sort -rn > /tmp/prose-files.txt || true
-            head -50 /tmp/prose-files.txt | awk '{print "| `" $2 "` | " $1 " |"}'
-            # No silent cap: say how many files the table leaves out.
-            more=$(( $(wc -l < /tmp/prose-files.txt) - 50 ))
-            if [ "$more" -gt 0 ]; then echo "| ... and $more more file(s) | |"; fi
-            echo
-            grep -E '^(PROSE GATE|prose gate):' /tmp/prose.txt || true
-          } >> "$GITHUB_STEP_SUMMARY"
-          exit $status
+          python scripts/quality_state.py prose --status "$status" --report /tmp/prose.txt >> "$GITHUB_STEP_SUMMARY"
       - name: Validate content tree
         run: python scripts/validate_content.py
       - name: Repo test suite (shape parity #1208, drift-gate units)

Makefile

@@ -1,6 +1,6 @@
 # Makefile für dein Adaptive-Learner-Content-Repo.
 #
-# Ein Befehl genuegt zum Loslegen:
+# Ein Befehl genügt zum Loslegen:
 #
 #     make validate        Prüft deine Inhalte (legt beim ersten Mal automatisch
 #                          eine lokale Python-Umgebung an, du musst nichts installieren).
@@ -40,7 +40,7 @@ PIP := $(VENV)/bin/pip
 ENGINE_PIN := $(shell cat schema/engine-version.txt)
 ENGINE_STAMP := node_modules/.engine-$(ENGINE_PIN)
 
-.PHONY: validate lint lint-warnings setup generate export export-anki audit clean help prose-check
+.PHONY: validate lint lint-warnings prose-check setup generate export export-anki audit clean help
 
 help:
 	@echo "make validate        - Inhalte prüfen (richtet sich beim ersten Mal selbst ein)"