Graph Definition: Rules, Traffic-Light Status, and Milestones (Appendix A)

Generated artifact. Do not edit manually.

Generated by: docs/concept/skill-graph/build_spec.py Regenerate with: cd app && npm run docs:skill-graph:web Source of truth: docs/concept/skill-graph/source/rules-and-gates.json

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

Compendium: §§1–5 Core · §§6–10 Context · §§11.1–11.4 Units · §11.5 Progression · §12 Planning · §13 Implicit Mastery · Appendices B–F · Appendix A (Rules & Ampel)

This appendix is the reference register for all 145 skill-graph rules (GD-*) in this document. For every rule, it answers three questions: what is required, which validation area owns the obligation, and what the existing checks actually prove.

How to read a row. Each GD-* row links an obligation to its designated validation area (target), existing checks (actual), and their evidentiary limits: ID (GD-, the document-part letter, and a sequential number: C for the formal core, X for context, U for learning units, P for planning, A for implicit-mastery recognition, and D for the document appendices); Traffic light (technical enforcement level on the mainline: HARD for rules enforced fail-closed across their entire declared domain, PARTIAL for rules hard-gated partially or via profiles or maturity floors, OPEN for binding obligations without an established hard gate on the core claim, and OPT/DOC for optional profiles, undecided policy, or document checks); Obligation (concise statement of the rule); Target (the validation area assigned to the obligation: a milestone M0–M7 or one of the area codes in Section A.2); Actual and gap (what existing checks prove and what remains open); Section (the named section where the rule is defined). Rows use concise telegraphic style and recurring shorthand: Unit for learning unit, Exam or Assessment for an exam or assessment goal, Memory for a memory goal or memorization, Normal and Strict for normal mode and strict mode, Closure for transitive closure, Scope for validation or applicability scope, Runtime for runtime execution, unit-wide lock for holistic unit gating, and human trial for learner evaluation.
Milestones and rule families. Milestones M0–M7 structure automated curriculum QA; runtime contracts, concepts, optional profiles, and artifact or document obligations are tracked separately. The two existing rule families are Curriculum Quality Review (CQR) and graph-validation rollout rule (GVR), denoting curriculum quality checks and skill-graph validator rollout rules respectively. The legend in Section A.2 defines every code referenced here along with its actual scope.
An assignment is not a pass. Target assignment names the validation area responsible for each obligation; the existing gates listed alongside prove only their actual scope. A binding rule without a complete gate is an open verification obligation, not an already passed check. M0 is a possible incomplete starting state, not a quality seal. Profile-bound route gates apply within their declared QA scopes. By contrast, the global base check for complete atomic normal form and transitive minimality (check code G17) must cover the entire declared canonical subject inventory, including relevant cross-subject prerequisites and inheriting ancestors; a QA profile subset or partial scan is insufficient.
Base before every milestone. Complete atomic normal form and a full transitive-minimality check are the required target base before any promotion to M1–M7. A G17 violation or an incomplete check blocks every milestone pass M1–M7 in the affected subject, even though individual source or review checks may still pass on their own. The atomic target state \(R \subseteq A \times A\) is binding: first establish or retain complete atomic prerequisites, next remove the non-atomic requirements rendered redundant by them, and only then transitively reduce the fixed atomic relation. An inherited alternative path alone never justifies deleting an atomic route. CQR-102 and GVR-012 remain directly atomic.
Memory goals. Flashcard repetitions belong to the learning process. Completing a memory goal requires its own explicit verification record; subsequent review due dates alone do not revoke that completion. Ordinary completed goals receive neither review due dates nor any mastery revocation derived from them. No new assessment entity or schema is introduced by this contract.
Scope of the register. The 145 stable GD-* IDs cover all six parts of the compendium. Section references use named sections rather than volatile line numbers. Runtime contracts, new recognition and learning-unit semantics, optional planning, and publication or document obligations are never assigned a fabricated M7 gate; where the Product Owner fixes a target area, such as M5/M7 for the coverage evidence of learning units (R2), the row states that it is not yet enforced. Subject-matter accuracy, human trials, and learning effectiveness remain independent forms of evidence.

Traffic-Light Status: What Is Truly Hard-Enforced Today

Many of the 145 GD-* rule groups bundle a technically gated structural core together with broader runtime, migration, or subject-matter obligations. The four-state traffic light therefore makes visible—both per individual validator gate and per rule group—where a technically hard (fail-closed) gate already operates today and where scope limits or open obligations remain.

Traffic Light	Meaning for Enforcement and CI	Count
🟢 HARD	Fully hard-enforced. The entire core claim is technically gated fail-closed across its declared domain; no material sub-obligation of the row remains open. Globally enforced in skill-graph, schema, UUID, and composition validators (`validate:graph`, `validate_schemas.py`, `validate_goal_ids_uuid.py`, `validate:composition-views`): JSON/schema, UUID format, resolvable references, self-reference protection, separate DAGs (`contains`, direct/inherited `requires`), `GVR-001/003/008/009/010/014/016`, and duplicate, single-parent, and non-emptiness rules for compiled views (`CPV-001..008`). Fully covered rule groups: GD-C05, GD-C07, GD-X07.	3
🟡 PARTIAL	Partially, profile-, or floor-bound hard-enforced. A named subset fails closed on violation, but applies only to protected subjects/profiles or covers only the structural or artifact share. Maturity floors (`check:curriculum-maturity-floors`) and status-file diff checks block regressions for ten Gymnasium subjects (5 at `M7`: Mathematics, Physics, Chemistry, Biology, Economics; 5 at `M6`: German, History, Computer Science, Latin, Politics and Economics); since 11 November 2026, `CQR-001` additionally applies the `G17` check (atomic normal form and transitive minimality) to every canonical Gymnasium root, hard via these floors (GD-C11, C15–C17, U13); `GVR-012/013` are hard-enforced in Mathematics, `GVR-006` in the Hesse rollout, and `validate:view-filters` (`APV-102/103`) for canonical Gymnasium views; complemented by the learning-scope closure compiler \(L\) (10 November 2026) as well as content-package, license, package, and release checks (54 groups).	54
🔴 OPEN	Binding, core obligation still without a hard gate. Normatively binding target, runtime, or concept rule without a cross-cutting fail-closed gate established in this register. Atomic expansion before closure in result derivation (GD-C13), open cluster atomization in `goalEntry` (GD-X09), cluster satisfaction across all atoms instead of `isCoreForPrereqs` (GD-C26), not-yet-implemented global implicit-mastery recognition including the learning-unit validity rules R1–R5 of 11 November 2026 (GD-A01–A18, C25, C30, X18, X32, X36), recursive `LearningUnit`s including their purpose and boundary criterion (GD-U01–U12, U15), and pure runtime, dialogue, and edge-provenance contracts (60 groups).	60
⚪ OPT/DOC	No production mandatory gate. Deliberately optional planning, explicitly undecided lock policy, or pure document/model verification outside curriculum and runtime gates. Optional acquisition and time planning (GD-P01–P11, X33–X35, X37), undecided unit-wide access lock (GD-U14), and appendix, model-checker, and PDF-build rules (GD-D01–D12; build and model scripts check fail-closed locally, while pure `docs/` changes are exempted in `ci.yml`; 28 groups**).	28

Total register: 3 HARD, 54 PARTIAL, 60 OPEN, 28 OPT/DOC (total: 145 rule groups).

Overview of Validation Areas

Code	Status	Meaning
`M0 Base, before M1`	🟡 PARTIAL	Obligation: Identity, references, structure, and base graph; complete atomic normal form and transitive-minimality check as the target base before every higher milestone pass; validity of every authored learning unit (R1, R3, R4). Actual limit: Schema and skill-graph validators check core invariants hard across all loaded skill landscapes; `CQR-002` checks explicit types of canonical Gymnasium roots; since 11 November 2026, `CQR-001` adds the full `G17` check for every root of the complete canonical Gymnasium catalog (a violation caps the subject at `M0`; hard in CI via the ten protected floors). Non-canonical roots have no `G17` gate; no unit validator (R1, R3, R4) exists.
`M1`	🟡 PARTIAL	Obligation: Readable, traceably registered source inventory. Actual limit: `CQR-000` hard-gated via the ten protected floors and status diff; does not yet prove subject-accurate mappings.
`M2`	🟡 PARTIAL	Obligation: Bidirectionally source-evidenced views, course-level mapping, and goal-count plausibility. Actual limit: `CQR-003/004/005` hard-gated via protected floors, each subject to its actual aggregator status logic.
`M3`	🟡 PARTIAL	Obligation: Additional complete directly atomic orientation-to-autonomy routes within the named scopes. Actual limit: `CQR-101/102/103` across twelve profiles; `GVR-012/013` only in Mathematics profiles. Does not fully gate the global target base.
`M4`	🟡 PARTIAL	Obligation: Visible route endpoints and concrete terminal assessment tasks. Actual limit: Actually `CQR-104/201/202` via protected floors; catalog target of `104` is `M3`. Does not prove live session grading conformance.
`M5`	🟡 PARTIAL	Obligation: Assessment approval/coverage metadata, current atomicity reviews, composition views, and cleared applicability warning debt; reviewed application coverage of every own ordinary unit goal (R2). Actual limit: Actually `CQR-203/301/401/501` plus composition and filter validators; catalog target of `203` is `M4`, and view file counting alone does not prove full domain sign-off. R2 is not checked: no units are authored.
`M6`	🟡 PARTIAL	Obligation: Justified memory-support decisions, traceable cards/decks, and required view visibility. Actual limit: `CQR-302` across all ten protected `M6`/`M7` floors plus card-review check; makes no claim about personal due dates or mastery revocation.
`M7`	🟡 PARTIAL	Obligation: Simultaneously current `D/P/A/M/V` evidence for all current `curricularAtomic` goals, all mandatory checks passed, and zero blockers; current per-goal binding of the R2 coverage evidence. Actual limit: `CQR-303` on five `M7` floors atop all lower stages; implies neither human sign-off, learner trials, learning effectiveness, nor general runtime conformance. None of the six required checks covers R2.
`Cross-Stage`	🟡 PARTIAL	Obligation: Learner state, satisfaction, frontier, Personal Curriculum, roles, focus, goal-type, and persistence contracts; versioned recognition contracts (R5). Actual limit: Closure compiler (10 November 2026), seed-scope fallback for non-compilable contexts (11 November 2026), and view/role checks established; remaining runtime and interaction proofs stay separate, with no blanket curriculum `M`-gate; R5 is a precondition before recognition is enabled.
`Concept / Optional`	🔴 OPEN / ⚪ OPT/DOC	Obligation: New global implicit-mastery recognition and recursive learning units (`OPEN`), plus optional time/budget planning and undecided unit access (`OPT/DOC`). Actual limit: Derive neither implemented new semantics nor a hard access lock from `M7`.
`Artifact / Document`	🟡 PARTIAL / ⚪ OPT/DOC	Obligation: Interchange, publication, licensing, data protection (`PARTIAL`), and document sign-off (`OPT/DOC`). Actual limit: Package, license, and release checks run in CI; a local document build is neither publication nor host acceptance.

Gate Reference and Actual Scope

Stable codes name individual existing checks (maturity and route documentation, graph validation rules). The legend explains every code; the subsequent assignment key maps their actual effect onto the milestones. The legend covers the exact eight GVR and eighteen CQR codes cited in this register; additional validator rules are neither introduced nor claimed as verified here.

Gate	Status	Actual Scope
`GVR-001`	🟢 HARD	A goal must not directly require any of its own `contains` ancestors; local skill landscape (all loaded skill landscapes).
`GVR-003`	🟢 HARD	A goal must not directly require its direct `contains` parent node; local skill landscape (all loaded skill landscapes).
`GVR-006`	🟡 PARTIAL	A goal must not directly require any of its direct `contains` children; named Hesse rollout inventory.
`GVR-009`	🟢 HARD	A stored structural node type must match its classification as leaf or cluster; local skill landscape (all loaded skill landscapes).
`GVR-010`	🟢 HARD	An optional `shortKey` must be unique within a logical skill landscape; language editions of the same goal may share the same ID/`shortKey` pair.
`GVR-012`	🟡 PARTIAL	Hard complete direct atomic route for every selected subject atom between same-stage orientation and terminal autonomy; canonical Mathematics lower and upper secondary separately.
`GVR-013`	🟡 PARTIAL	Direct local prerequisites of selected non-memory atomic sources must target atomic goals; configured Mathematics profiles only, not a global edge check.
`GVR-014`	🟢 HARD	Orientation goals are prerequisite-free atoms without exam, memory, example, or assessment metadata; all loaded skill landscapes.
`CQR-000`	🟡 PARTIAL	Source inventories are readable, linked to official HTTP(S) sources, and have their extracted goals registered.
`CQR-001`	🟡 PARTIAL	Since 11 November 2026: every declared root of the complete current canonical Gymnasium catalog, with its required cross-subject goals and inheriting ancestors, passes the shared reference and effective-DAG audit, atomic `requires` normal form, explicit type consistency, and transitive minimality (`G17`).
`CQR-002`	🟡 PARTIAL	Stored structural node types match the atomic or cluster structure.
`CQR-003`	🟡 PARTIAL	Jurisdiction views have atomic source evidence and cover all registered original source goals completely; checked bidirectionally.
`CQR-004`	🟡 PARTIAL	Upper-secondary source goals match the `GK`/`LK` mapping of canonical goals; unspecified course levels apply to both `GK` and `LK` unless a reviewed `LK` exception exists.
`CQR-005`	🟡 PARTIAL	Extracted goal counts are plausible against the reviewed `HE`/`BW` baseline or have explicitly reviewed deviations.
`CQR-101`	🟡 PARTIAL	Complete routes across effective (including inherited) prerequisites connect motivation and terminal autonomy; configured QA scopes.
`CQR-102`	🟡 PARTIAL	Complete routes across directly authored atomic prerequisites connect motivation and terminal autonomy; configured QA scopes.
`CQR-103`	🟡 PARTIAL	Clusters selected within the QA scope carry no `requires`; not a global proof across all authored edges.
`CQR-104`	🟡 PARTIAL	Motivation and terminal autonomy endpoints are visible in the relevant learner-facing composition views; scopes configured accordingly.
`CQR-201`	🟡 PARTIAL	Terminal autonomy goals of the configured scopes carry `examData`.
`CQR-202`	🟡 PARTIAL	Their exam tasks are concrete and usable in assessment mode rather than placeholders.
`CQR-203`	🟡 PARTIAL	Their assessment metadata specifies approval status, covered goals, content areas, and cognitive demand level.
`CQR-301`	🟡 PARTIAL	Configured atomicity reviews are complete, current, and free of open review findings.
`CQR-302`	🟡 PARTIAL	Memory-support decisions, required flashcards, and decks are traceable; required memory goals are visible in the configured views.
`CQR-303`	🟡 PARTIAL	For a non-empty current inventory of `curricularAtomic` goals, `D/P/A/M/V` evidence is jointly present, all mandatory checks pass, and zero blockers exist; automated QA, not human sign-off.
`CQR-401`	🟡 PARTIAL	Composition-view files exist; actual file counting does not prove complete domain sign-off of view contents.
`CQR-501`	🟡 PARTIAL	Active applicability warnings are cleared and accepted warning waivers match the current findings.

Assignment Key

Register rows name validation areas using the eleven codes below. Each entry explains a code together with its existing checks and their evidentiary limits.

BASE. CQR-001 (references, effective DAG, atomic normal form, explicit types, and transitive minimality, i.e. G17, over the complete canonical Gymnasium catalog since 11 November 2026) and CQR-002 (explicit structural node types) carry catalog target M0. validate:graph runs additional standalone GVR and integrity checks across all loaded skill landscapes. Unit validity (R1, R3, R4) is assigned here but has no validator yet. This does not imply complete verification of every mathematical or runtime rule.
SOURCE. CQR-000 \(\to\) M1; CQR-003/004/005 \(\to\) catalog target M2. Covers source inventory, view coverage, course-level mapping, and count plausibility; the concrete aggregator status logic is not equally strict across all checks. Subject-matter accuracy does not follow from technical evidence checks alone.
ROUTE. CQR-101/102/103 \(\to\) M3, configured QA scopes only. CQR-102 uses exclusively directly authored atomic edges; CQR-103 forbids requires on selected clusters. GVR-012/013 are additional scoped profiles, not a global completeness proof. CQR-104 carries catalog target M3 but blocks only M4 in the actual curriculum promotion chain.
EXAM. CQR-201/202/203 carry catalog target M4. 201/202 are actual prerequisites for curriculum M4; 203 takes effect only together with core reviews at M5. Metadata and task-readiness checks prove neither accurate live-session grading nor independent application coverage across every passing score combination.
VIEW. CQR-401 and CQR-501 \(\to\) M5. 401 counts existing view files rather than proving their complete domain sign-off. Standalone validate:composition-views and validate:view-filters gates check further, individually bounded structural and projection obligations. Never equate these blanketly with all runtime rules.
REVIEW. CQR-301 \(\to\) M5 (current semantic atomicity decisions), CQR-302 \(\to\) M6 (memory decisions, cards, and visibility), CQR-303 \(\to\) M7 (complete current D/P/A/M/V intersection with required checks passed and zero blockers). M7 remains automated curriculum QA.
RT. Cross-stage runtime and interaction contract; no curriculum M-gate is designated for this overall claim. Individual implementations and tests must be evidenced separately; their mere existence is not a live conformance check.
CONCEPT. Normatively specified, but new global implicit-mastery recognition and recursive learning-unit semantics are explicitly not yet implemented in production. No existing curriculum M-gate. Structural or content shares may be mapped to target validation areas for future implementation without fabricating M7 fulfillment.
OPTIONAL. Optional planning profile or still-undecided holistic unit-access policy. No curriculum M-gate; do not derive a new hard runtime lock from the specification.
ARTIFACT. Standalone publication, interchange, licensing, or data-protection contract; no automatic M0–M7 assignment. Any specific package or release check must be evidenced within its own declared scope.
DOCUMENT. Document build, mathematical formula model, or diagram check; neither a curriculum M-gate nor proof of runtime conformance or learning effectiveness.

Actual Promotion Chain and Additional Gates

Actual maturity promotion is governed by deriveScopeMaturity and deriveCurriculumMaturity in the status generator; catalog metadata alone never computes maturity. The actual promotion chain requires CQR-104/201/202 only at curriculum M4 and CQR-203 together with core reviews at M5; distinguish these from the catalog targets of 104 (M3) and 203 (M4).

M0 \(\to\) M1: CQR-001/002 and CQR-000 must be pass. Both base graph rules carry catalog target M0; since 11 November 2026, CQR-001 includes the G17 check, whose violation also caps every scope at M0.
M1 \(\to\) M2: CQR-003 must be pass; CQR-004 blocks on fail, CQR-005 on warn. not_configured on either of those two rules does not block independently in the aggregator.
M2 \(\to\) M3: Route profiles must exist; CQR-101/102/103 must be pass in every configured scope.
M3 \(\to\) M4: Configured CQR-104 and CQR-201/202 must be pass in every scope. CQR-104 carries catalog target M3.
M4 \(\to\) M5: All scoped CQR-203 checks plus CQR-301/401/501 must be pass. CQR-203 carries catalog target M4; CQR-401 counts registered view files.
M5 \(\to\) M6: Additionally CQR-302 must be pass.
M6 \(\to\) M7: Additionally CQR-303 must be pass: non-empty current curricularAtomic denominator, complete D/P/A/M/V intersection, all six mandatory checks passed, and zero blockers.
Missing route profiles cap an otherwise source-ready curriculum at M2. CQR-004 blocks in the aggregator on fail and CQR-005 on warn; not_configured on either does not block independently there. CQR-401 counts view files. Source coverage binds existing parseable registered snapshots rather than proving exhaustive extraction of official PDFs on its own.
Routes are evaluated across twelve explicit profiles for ten curricula: Mathematics and Physics each in lower and upper secondary separately; Chemistry and Biology each in upper secondary; Economics, Politics and Economics, Computer Science, History, German, and Latin each cross-stage. GVR-012/013 apply only within their Mathematics profiles (GVR-012 remains an error even in legacy warn mode); GVR-013 covers direct local atomic sources for non-memory atoms rather than all subjects, cluster sources, or cross-subject edges. CQR-103 checks only the selected clusters. The atomic target state \(R \subseteq A \times A\) is gated by CQR-001 (canonical Gymnasium catalog only, hard via floors); the ordered migration is evidenced only for reductions bound to the prerequisite-reduction proof registry.
The transitive-minimality auditor also runs standalone with CLI and contract tests (check mode fails on findings, whereas report mode can emit findings with exit code 0); CI runs its tests (test:transitive-minimality). Since 11 November 2026, its zero-finding result per canonical Gymnasium root, together with atomic normal form, is part of CQR-001 through the graph-integrity adapter; wherever CQR-001 passes, the historical raw-edge criterion coincides with atomic minimality. Before emitting status rows, the generator verifies the prerequisite-reduction proof registry against the complete current canonical catalog and aborts on any deviation.
Commands quality:curriculum-status, quality:curriculum-status:check, and check:curriculum-maturity-floors are registered in package.json. The status producer computes and writes the report; domain warnings or failures lower maturity without changing its exit code on their own. The CI workflow then runs the floor checker against the floor policy and verifies a clean git diff of generated status files. Floor checking trusts the numerically computed maturity on a single matching subject entry and computes no independent CQR/G17 verdict. Five curricula are protected at M7 and five at M6; experimental inventories without a floor remain reportable at M0. Standalone graph, route, source, filter, and composition checks act as additional CI gates rather than extra CQR rules.

Workflow triggers are manual dispatch, push to main, and pull request to main under path filters; pure documentation changes are generally exempted while generated QA status files are re-included. No single CI or build result on its own proves all runtime, subject-matter, and learning-effectiveness rules.

C – Formal Core

Sections 1–5 (01-core). Every row cites its section; GD-* IDs remain stable regardless of later section renumbering.

ID	Status	Section	Rule (concise summary)	Area	Enforcement and boundary
`GD-C01`	🟡 PARTIAL	§1.1	Finite learning-goal graph \((G,C,R)\); `contains` and `requires` are the sole authored goal edges. Inverse and non-empty transitive closure are distinct operations. DAG checks apply to the respective named relations rather than blanketly to \(C \cup R\).	`BASE` (M0 base); semantic definition cross-stage	`BASE`, `GVR-008`, and schema enforce finite goal sets, permitted edge fields, and separate DAGs; mathematical operation semantics is not a standalone `CQR` pass.
`GD-C02`	🟡 PARTIAL	§1.1	`contains`: parent goal \(\to\) child; `requires`: dependent goal \(\to\) prerequisite. Any learning-flow diagram drawn in the reverse direction must explicitly declare the inversion.	`BASE` (M0 base) / presentation cross-stage	`BASE`/`ROUTE` consume the edge direction; diagram labeling is not a `CQR` check.
`GD-C03`	🟡 PARTIAL	§1.2	Canonical UUID identity is injective and persists across language, placement, and view; local formula letters do not define an alternative storage format.	`BASE` (M0 base); identity preservation additionally `RT`/`ARTIFACT`	`BASE` plus UUID and view validators enforce ID format and uniqueness, but do not on their own prove all version and migration guarantees.
`GD-C04`	🟡 PARTIAL	§1.2	Base attributes are total: ID, title, and strictly positive weight. `description` must be intelligible and verifiable; `examples` are illustrative references, neither quality proof nor competency evidence.	`BASE` (M0 base) for form/data domain; M7 for current content evidence	Schema enforces required fields and `REVIEW` (on M7 floors) enforces current content evidence; strictly positive weight and prose quality are not globally gated across all inventories.
`GD-C05`	🟢 HARD	§1.2	An optional `shortKey` is unique per logical skill landscape; localized language files for the same goal may reuse the same ID/`shortKey` pair.	`BASE` (M0 base)	Hard-enforced via `GVR-010` in `validate:graph` across all loaded skill landscapes, including the localization boundary.
`GD-C06`	🟡 PARTIAL	§1.2	`phase` is compatibility/display metadata, neither identity nor sequencing. `area`/`tags` never replace authoritative goal-type or scope decisions; `level` denotes difficulty; `core` does not on its own determine the selected target scope; `weight` denotes progress share rather than duration or evidentiary strength; new optional materials belong in content packages.	Cross-stage model contract; M2/M5 for respective mapping/view evidence; material contract `ARTIFACT`	Content-package validator and `GVR-008` enforce material separation; no single gate checks every field's semantic boundary. For sources and materials see `GD-X38`–`GD-X43`.
`GD-C07`	🟢 HARD	§1.3	Structurally atomic iff a goal has no `contains` children; otherwise a cluster. Polyhierarchy is permitted; explicit structural types must match. Pedagogical goal kinds are neither new structural types nor program units.	`BASE` (M0 base)	Structural type, leaf/cluster classification, polyhierarchy, and separation from program units are hard-enforced globally via `GVR-009`, `CQR-002`, and schema; semantic atomicity is governed separately by `GD-C08`.
`GD-C08`	🟡 PARTIAL	§1.3	A structural leaf is not automatically semantically atomic; bundling multiple independent sub-competencies can fail content review.	M5, deeper current binding M7 (`REVIEW`)	Hard-protected via `CQR-301` and `CQR-303` across the ten M6/M7 floors; unprotected inventories remain below, and structural leaf status alone is insufficient.
`GD-C09`	🟡 PARTIAL	§2.1	`Anc`/`Desc` follow non-empty `contains` paths; \(\Atoms(a)=\{a\}\) for atoms and atomic descendants otherwise. In a finite DAG every cluster has leaf descendants; set semantics counts multi-paths once. Childless means atom, never empty cluster.	`BASE` (M0 base); `RT` for evaluation	`BASE` proves structural invariants, not every runtime aggregation path.
`GD-C10`	🟡 PARTIAL	§2.2	Multiple prerequisites apply conjunctively. Alternatives require an explicit selection contract; \(R\) is acyclic.	`BASE` (M0 base); selection contract separate	`BASE` checks DAG acyclicity; there is no tacit OR semantics.
`GD-C11`	🟡 PARTIAL	§2.2; §2.3	In the target state, all domain prerequisites are authored strictly between atomic learning goals: \(R \subseteq A \times A\). Cluster inheritance remains a transitional compatibility mechanism and never substitutes for complete atomic modeling. Never delete an atomic route in favor of an inherited cluster representation.	`BASE` (M0 target base, before M1): necessary prerequisite for the full `G17` check; never deferred to M3	Since 11 November 2026, `CQR-001` reports every authored requirement with a non-atomic source or prerequisite for every canonical Gymnasium root and its required closure; a finding caps the subject at `M0`. CI fails on it only via the protected floors; other canonical subjects report `M0`, and non-canonical roots are not checked. `ROUTE` additionally checks scoped profiles.
`GD-C12`	🟡 PARTIAL	§2.3	Effective direct requirements equal a goal's own requirements united with those inherited from all `contains` ancestors; multiple parents yield set union. Inheritance and transitive closure are distinct operations.	`BASE` (M0 base for valid transition inventory); `RT`	`validate:graph` checks effective cycles; that alone does not prove all runtime evaluations.
`GD-C13`	🔴 OPEN	§2.3	Expand effective requirements to atoms first, then form the transitive closure; further prerequisites of expanded atoms remain included. Mastering an intermediate goal never truncates the closure.	`BASE` (M0 target base for normalized validity); `RT`/`CONCEPT` for result derivation	Not blanketly covered by `CQR-001` or historical `G17`/completion tests.
`GD-C14`	🟡 PARTIAL	§3.1	`contains`, direct `requires`, and effective `requires` are acyclic; self- or back-dependencies induced via atomic expansion are likewise invalid.	`BASE` (M0 base)	`BASE` and effective DAG checks hold only for the relations they actually evaluate; full atomic expansion must be verified separately.
`GD-C15`	🟡 PARTIAL	§3.1	Requiring the same target both directly and via ancestor inheritance is locally redundant. During migration, removing this duplication must never roll back the complete atomic representation.	`BASE` (M0 target base, before M1, jointly with atomic normal form)	`CQR-001` flags redundant authored requirements, including duplicates of inherited ones, and non-atomic requirements in canonical Gymnasium roots (hard via protected floors); delivered removals are bound to the replayed prerequisite-reduction proof registry. Non-canonical roots are not checked.
`GD-C16`	🟡 PARTIAL	§3.1	Transitive minimality is binding. First establish or retain complete atomic prerequisites, next remove redundant non-atomic requirements, and only then transitively reduce the atomic relation. Testing raw authored edges against an inherited alternative path alone never justifies deleting an atomic edge.	`BASE` (M0 target base, before M1): full `G17` check on the atomic target state; any violation or incomplete check blocks every subject milestone pass M1–M7	Since 11 November 2026, `CQR-001` runs the full `G17` check (atomic normal form and transitive minimality, including required cross-subject goals and inheriting ancestors) for every root of the complete canonical Gymnasium catalog; a violation caps the subject at `M0` and so blocks `M1`–`M7`, and an incomplete catalog aborts the status run. Hard in CI via the protected floors only; other canonical subjects report `M0`, and non-canonical roots are not checked. A pass certifies the resulting state; the migration order is evidenced only for reductions bound to the proof registry. `CQR-102`/`GVR-012` remain directly atomic.
`GD-C17`	🟡 PARTIAL	§3.1	Transitively reducing a fixed atomic relation preserves its closure. When modifying a mixed inventory, recompute inheritance and expansion; preserve atomic demand and all relevant scope, applicability, and frontier contracts. Raw reachability between non-atomic cluster nodes is not a general preservation invariant.	Cross-stage migration contract; fulfilling it underpins the atomic `G17` target state as `BASE` (M0 target base before M1)	Registered prerequisite reductions of the canonical Gymnasium catalog are replayed with their witnesses against digest-bound baseline and current documents in every status run, which aborts on failure; this covers the closure share of delivered reductions. It does not prove preservation of the further scope, applicability, and frontier contracts in general, nor anything for non-canonical roots; a raw-versus-effective reachability comparison on its own does not prove a valid migration.
`GD-C18`	🟡 PARTIAL	§3.1	Identities and references must be resolvable; unknown goals are never treated as satisfied prerequisites.	`BASE` (M0 base) + `RT`	`BASE` checks static references; fail-closed runtime semantics is verified separately.
`GD-C19`	🟡 PARTIAL	§3.2	Single-parent tree structure is an optional profile rather than a global DAG requirement. Forbidding own ancestors is profile-bound; requiring own descendants can already induce effective self-dependency.	`BASE` (M0 base for actual cycles); optional profile separate	`GVR-001/003/006` operate within declared scopes; there is no universal single-parent tree mandate.
`GD-C20`	🟡 PARTIAL	§3.2	Skill-graph validity, projection validity, and full learnability from a declared entry state are distinct claims. Atomic self-dependency and an empty initial frontier can refute learnability.	`BASE` (M0 base) + `VIEW`; learnability profile separate	A single green validator pass never proves all three claims simultaneously.
`GD-C21`	🟡 PARTIAL	§3.3	Every non-exempt atom lies on a direct atomic route Orientation \(\to\) Atom \(\to\) Terminal Autonomy. Exemptions require a documented machine-readable selection contract; clusters do not serve as route bridges.	M3 (`ROUTE`)	`ROUTE`; `GVR-012` applies only within its declared profile. The formula is a structural criterion, not a proof of pedagogical quality.
`GD-C22`	🟡 PARTIAL	§3.3	Route profiles define stage, goal kinds, ordering, and validation scope. Memory support is added only where needed, not as a universally mandatory intermediate step.	M3 (`ROUTE`); memory content decision M6 (`REVIEW`)	`CQR-101/102/103` and `GVR-012/013` apply within scoped profiles; `CQR-302` is checked separately.
`GD-C23`	🔴 OPEN	§3.3	Cluster overviews may be derived from atomic reachability. An existential summary edge is never a universal prerequisite; percentage coverage requires a declared denominator; author no new cluster `requires`.	Cross-stage presentation; prohibition of new authored cluster edges follows from `BASE` (M0 target base before M1)	No standalone `CQR` pass covers overview display semantics.
`GD-C24`	🔴 OPEN	§4.1	Exactly one stored binary state per atom; partial scoring creates no third mastery tier. Ordinary completions evidence understanding, whereas special goal types use their own completion contracts.	`RT`, cross-stage	No curriculum `M`-gate; see goal-type rules.
`GD-C25`	🔴 OPEN	§4.1	\(\Meff\) equals explicit mastery united with derivations from durably valid local contracts. It acts globally on prerequisites and actual progress across focus switches without writing synthetic mastery records. Orientation denotes only a completed entry.	`CONCEPT`, no M-gate	New global implicit-mastery recognition is explicitly not yet implemented; M7 does not prove it.
`GD-C26`	🔴 OPEN	§4.1	\(\Sat(g)\) holds iff \(g\) has a non-empty atomic basis contained completely in \(\Meff\). Mastering a dependent content goal alone never proves its prerequisites, and cluster completion never writes mastery back to children.	`RT`/`CONCEPT`	Appendix B documents the historical `core`-biased cluster aggregation divergence; live conformance must be verified separately.
`GD-C27`	🔴 OPEN	§4.2	The global frontier consists of unmastered atoms whose full transitive effective atomic prerequisites are satisfied; traversal never stops at mastered intermediate goals, and partial work never counts as satisfaction.	`RT`/`CONCEPT`	No curriculum `M`-gate covers this overall runtime claim.
`GD-C28`	🔴 OPEN	§4.2	A separate cluster frontier uses own and inherited atomic bases together with transitive closure and never replaces the atomic learning frontier. The UI may apply further explicit selection rules.	`RT`	Never equate the mathematical frontier with every UI menu listing.
`GD-C29`	🟡 PARTIAL	§4.3	Weighted progress counts unique target atoms rather than clusters and children simultaneously. Weights must be positive; an empty target set never receives a fabricated 100% value. Uniform weights are a domain convention rather than a requirement; the Personal Curriculum fixes the target denominator. Progress share is neither fractional per-goal mastery nor a grade or probability.	`RT`; attribute base `BASE` (M0)	Schema enforces numeric `weight` and the closure compiler (10 November 2026) fixes the personal denominator \(L\); no global gate checks strictly positive weights and every display formula.
`GD-C30`	🔴 OPEN	§4.3	Segmented progress views may distinguish explicit and implicit contributions, while total progress counts their union. Any counter restricted to explicit mastery must be labeled as such.	`RT`/`CONCEPT`	Joint explicit/implicit progress accounting is proven neither by M7 nor by current display components.
`GD-C31`	🟡 PARTIAL	§5.1	A filter is a predicate; applicability is a partial dimensional map. An omitted dimension is unrestricted, an explicit empty value set excludes concrete values, and `ALL` ignores a dimension solely as a query value and is never persisted. An unresolved personal stage stays unresolved.	`VIEW`/`RT`; M2/M5 only for respective source/warning-debt shares	Filter validator and `CQR-501` check compiled applicability in canonical scopes; not a blanket proof of every runtime filter case.
`GD-C32`	🟡 PARTIAL	§5.1	Applicability overrides justify exceptions; runtime evaluates compiled applicability. Visibility alone never proves official source coverage.	M2 (`SOURCE`) for source evidence; M5 (`VIEW`) for warning debt; `RT`	`SOURCE`/`VIEW` are checked separately; an accepted override waiver never substitutes for `CQR-003` source coverage.
`GD-C33`	🟡 PARTIAL	§5.2	Preserve the full transitive atomic closure prior to view or focus restriction; global inheritance and atomic expansion always precede filtering. Under the 10 November 2026 Product Owner decision, this closure together with the selected seed targets forms the fixed personal learning scope \(L\). Hidden children never turn a cluster into an atom.	`VIEW`/`RT`	Verify full closure and identity preservation separately; existing opaque `goalEntry` cluster atomization remains an open implementation gap.
`GD-C34`	🔴 OPEN	§5.3	Normal mode restricts globally computed transitive atomic requirements to focus endpoints; an empty relative prerequisite basis does not block. Strict mode checks all global prerequisites for candidates inside the focus. The strict frontier is a subset of the normal frontier. Out-of-focus prerequisites are never marked mastered.	`RT`/`CONCEPT`	No curriculum `M`-gate. Global closure computation prior to filtering must be verified separately.
`GD-C35`	🔴 OPEN	§5.3	Given \(g \to x \to p\) and focus \(\{g,p\}\), \(p\) remains required in normal mode while strict mode additionally requires \(x\). Mastering \(x\) never removes the transitive requirement on \(p\).	`RT`/`CONCEPT`	Distinct mediated-transitivity case, not proven by direct-neighbor checks alone.
`GD-C36`	🔴 OPEN	§5.4	Global diagnostics reports all missing transitive atoms and distinguishes in-focus from out-of-focus blockers. Listing direct clusters is insufficient; diagnostics creates neither edges nor mastery and is distinct from normal-mode blockers.	`RT`	No curriculum `M`-gate.

X – Context, Views, Goal Types, Provenance, and Quality

Sections 6–10 (02-context). Every row cites its section; GD-* IDs remain stable regardless of later section renumbering.

ID	Status	Section	Rule (concise summary)	Area	Enforcement and boundary
`GD-X01`	🟡 PARTIAL	§6	Goals, program units, placements, competency axes, and composition views serve distinct responsibilities. Reordering creates no new competency and deletes no learner results; broad taxonomy categories are not concrete assessable skills.	Cross-stage model contract; M5 target (`VIEW`) for delivered views	Schema and composition validator separate the layers structurally; complete migration across all legacy inventories is not claimed.
`GD-X02`	🟡 PARTIAL	§6.1	Goal identity persists across new placements. Program-unit type `exam` is not an exam goal; `phase` induces neither `contains` nor `requires` edges. Historical year trees are compatibility scaffolding rather than durable domain identity.	`BASE` (M0 base for identity); `VIEW`/`RT` for projection	`BASE`/`VIEW` cover only structural sub-obligations.
`GD-X03`	🟡 PARTIAL	§6.2	`programUnits` form a tree or forest; `goalPlacements` distinguish `primary`, `secondary`, and `assessed` relations. Secondary relevance or assessment links never create duplicate visible goal nodes.	M5 target / `VIEW`	Schema and composition validator enforce structure and duplicate-freeness of compiled views; `CQR-401` on its own only counts files.
`GD-X04`	🟡 PARTIAL	§6.2	Within a resolved context, a goal has at most one matching primary anchor; an absent placement never permits inventing a parent node. Multiple matching primary placements require reviewed resolution. Context never alters goal IDs.	M5 target / `VIEW`	Placement and projection checks run separately; they imply no automatic domain review.
`GD-X05`	🔴 OPEN	§6.2	Content view follows `contains`, program view follows program units and placements, and competency view follows catalog/competency references. Child order stems from declared data order; topological learning order is a distinct view.	`VIEW`/`RT`	No blanket `M`-gate covers all view-rendering semantics.
`GD-X06`	🟡 PARTIAL	§6.3	Composition views specify tree structure, child order, and canonical references—never newly authored atoms or `requires`. Renderers must not override an authoritative view with title, phase, or placement heuristics.	M5 target (`VIEW`); `RT`	Composition validator enforces reference-based views free of authored atoms or `requires`; the UI renderer contract remains separate.
`GD-X07`	🟢 HARD	§6.3	The compiled default learner tree displays each goal at most once and under at most one parent node. Canonical polyhierarchy requires deterministically validated resolution; role resolution alone never legitimizes duplicates.	M5 target / `VIEW`	Hard-enforced across all composition views via `CPV-004/005/006` in the composition validator; imposes no global single-parent constraint on canonical goals.
`GD-X08`	🟡 PARTIAL	§6.3	Compile early: select context/view, resolve references and full prerequisite closure, verify explicit roles and visible parents, and validate the compiled tree prior to UI rendering.	M5 target / `VIEW`	Composition, role, and filter checks verify static compilation of canonical views; `CQR-401` only counts files, and cross-subject `APV-102` alignment remains open.
`GD-X09`	🔴 OPEN	§6.3	No canonical cluster may be turned into an independently completable surrogate atom via `goalEntry` or display compaction; this rule holds regardless of any semantic-kind review. The canonical atomic basis and required prerequisite chains remain intact, with neither new view `requires` nor mastery propagation to children.	M5 target / `VIEW`	`CPV-009` blocks only `curricularArea` nodes in views; existing runtime cluster atomization is not yet remediated, and M7 does not prove it.
`GD-X10`	🔴 OPEN	§7.1	Catalog navigation precedes personalization. Level 1 is the base curriculum, Level 2 the backend-owned durable personal scope, Level 3 the temporary focus with one active atom, and Level 4 the binary result bound to stable goal IDs. View or focus changes preserve results; cluster state is derived.	`RT`, no curriculum `M`-gate	Curriculum QA maturity implies no blanket runtime implementation claim.
`GD-X11`	🔴 OPEN	§7.1	`LK` is a per-subject course attribute rather than an implicit stage; an absent stage remains unresolved. Changing Level 2 revalidates focus and the active goal without granting or deleting global results. Configuration occurs in designated first-party SkillPilot surfaces, never autonomously inside the OpenAI V1 coach.	`RT`	Provider and authorization boundaries are not an M7 gate.
`GD-X12`	🟡 PARTIAL	§7.1	\(L\) comprises seed targets together with all necessary atomic prerequisites, independent of current mastery; mastered goals stay satisfied inside the denominator. Focus narrows active work only, never total progress or completion. Open necessary atoms require an acquisition target in the full Personal Curriculum; each canonical ID counts once.	`VIEW`/`RT`; M2/M5 only for concrete view shares	10 November 2026 Product Owner decision; implemented in backend and WebGUI. Full native, curriculum, and remote CI acceptance remain separate; `CQR` maturity is no substitute.
`GD-X13`	🟡 PARTIAL	§7.2	Authored subject roles and their specificity precedence hold: omitted role defaults to `target`, direct `goalEntry` overrides subtree inheritance, closer root overrides broader root, and ties resolve to `target`. Roles are never inferred from age or phase. The explicit personal closure rule adds necessary atoms; a subject-local `prerequisiteOnly` role is valid when an acquisition target exists elsewhere in \(L\).	`VIEW`/`RT`	`CPV-008`, role unit tests, and the closure compiler check `projectionRole` and specificity; no single `CQR` gate proves every cross-view overlap case.
`GD-X14`	🟡 PARTIAL	§7.2	Authored seed references remain hard-validated. A necessary atom's compiled target role backed by a verified canonical path and provenance suffices without manual per-atom authoring. Unknown references, cycles, and missing or unreachable necessary atoms are errors of validation, curriculum QA, and native acceptance; under the 11 November 2026 decision, a learner read of a context whose canonical graph is not strictly compilable keeps the reviewed seed scope without derived closure, claims no learning basis, offers no goal with broken demand, and logs once per curriculum and error code. Seed-selection errors stay hard. Normal mode continues to ignore valid out-of-focus requirements.	Cross-stage projection/`RT` contract	`CPV-002` and `validate:graph` enforce reference and cycle errors; the personal closure (10 November 2026) and the seed-scope fallback (11 November 2026, `b511654efb`, with regression tests) are implemented in the backend; applicability and migration acceptance remain separate.
`GD-X15`	🟡 PARTIAL	§7.3	Applicability describes evidenced visibility and placement describes structural location; neither implies mastery. Declared mapping-inheritance boundaries shield supplementary branches from false broad provenance while retaining direct sources. Local endpoint visibility may follow from `requires` under an explicit contract without proving official source coverage.	M2 (`SOURCE`) for source evidence; M5 (`VIEW`) for warning debt/views; `RT`	`SOURCE`/`VIEW`; inheritance boundaries and derived applicability never automatically satisfy `CQR-003`.
`GD-X16`	🔴 OPEN	§7.4	Focus is the union of selected visible roots and their descendants within the Personal Curriculum. Normal mode checks internal transitive effective requirements while strict mode checks global ones; ignoring out-of-focus blockers never grants mastery.	`RT`	Refers to `GD-C33`–`GD-C35`; no additional curriculum `M`-gate.
`GD-X17`	🔴 OPEN	§7.4	Focus expansion follows the unique visible parent path, forms a strict superset containing at least one open goal, replaces only subsumed multi-root branches while preserving independent roots, and requires the coach to adopt the exact backend option and await learner confirmation.	`RT`	Standalone backend and interaction conformance; not an `M`-gate.
`GD-X18`	🔴 OPEN	§7.4	Only genuine focus completion justifies expansion; an empty frontier may indicate a prerequisite blockade. Newly exposed prerequisites are evaluated normally; mere successor mastery never propagates backward, whereas validly recognized implicit application persists globally and is never reopened.	`RT`/`CONCEPT`	Global recognition remains conceptual; distinguishing true completion from blocked empty frontiers is verified separately.
`GD-X19`	🟡 PARTIAL	§8	The skill graph receives a binary result resolved according to goal type and never grades answers, dialogues, or flashcards itself. Ordinary understanding must cover all required competency aspects; one right exercise is not automatically full mastery, nor is there an arbitrary minimum task count.	`RT`; positive content evidence M7 (`REVIEW`)	`CQR-303`/`P` checks authored performance profiles, not live learner success per session.
`GD-X20`	🔴 OPEN	§8.1	Sufficiently evidenced mastery is persisted immediately; success may be announced only after confirmed persistence. Extra learner consent to save is not a completion condition. Moving to a new topic requires a backend-selected proposal and recognizable learner agreement.	`RT`	Persistence and coach-interaction contract; not an `M`-gate.
`GD-X21`	🟡 PARTIAL	§8.2	Orientation conveys honest, motivating possibilities and tests neither prior knowledge nor subject competency, recall, or transfer; there are no motivation percentages. Orientation nodes are pure prerequisite-free atomic entries without exam or memory metadata.	`BASE` (M0 base)/M3 (`ROUTE`) for structure; `RT` for dialogue	`GVR-014` enforces structure and route profiles check embedding; dialogue quality is not proven thereby.
`GD-X22`	🔴 OPEN	§8.2	Selecting an interest opens exploration rather than immediate completion: concrete connection, personal response, then—after a meaningful reply or explicit wish to move on—a positive summary with a separate completion offer, completed only upon further confirmation. Never apply this flow to content goals.	`RT`	No curriculum `M`-gate covers the conversational turn sequence.
`GD-X23`	🔴 OPEN	§8.2	Upon a genuine fresh start without subject progress, open orientation goals are the sole selectable frontier goals; existing subject progress never reverts a learner back into an entry lock.	`RT`	`GVR` route and orientation structure checks do not prove personalized frontier entry logic.
`GD-X24`	🟡 PARTIAL	§8.3	A memory goal bundles its required flashcards. SRS practice and controlled verified recall are distinct process states; card flipping or self-rating is never proof of verified recall. Exactly one binary goal result exists, with no extra mastery tiers.	`RT`; M6 (`REVIEW`) only for memory content justification	`CQR-302` does not inspect individual learner SRS or recall results.
`GD-X25`	🔴 OPEN	§8.3	Repetitions belong to flashcard learning; completing a memory goal requires its own explicit verification record. Subsequent review due dates alone do not revoke that completion. Ordinary completed goals receive neither review due dates nor any mastery revocation derived from them; voluntary practice stays separate.	`RT`, binding Product Owner decision; no curriculum `M`-gate	Individual practice scheduling and durable goal completion are separated. Historical code descriptions are dated audit findings rather than competing norms. The contract introduces neither an extra assessment entity nor a new schema; specification prose is not proof of runtime implementation.
`GD-X26`	🟡 PARTIAL	§8.3	Every required flashcard deck must resolve completely and validly; missing cards must never be treated as an empty auto-completed deck.	M6 (`REVIEW`) for authored deck integrity; `RT` for individual completion	Memory-card review checks and `CQR-302` enforce authored deck integrity in CI and on M6/M7 floors; individual runtime completion stays separate.
`GD-X27`	🔴 OPEN	§8.3	The specified SRS process tracks interval, repetition count, ease factor, and next due date: unscheduled cards are due immediately, successful practice extends intervals, and insufficient recall resets the streak. Scheduling never substitutes for verified recall.	`RT`	Documented runtime process; M6 proves neither its runtime conformance nor any mastery revocation.
`GD-X28`	🔴 OPEN	§8.3	Controlled recall may begin on the active memory goal even while SRS practice items remain queued: prompt first, independent learner response second, expected answer and recall evaluation third. Passing updates the practice trace; failure affects the card process.	`RT`	Card scheduling and goal completion are separate claims. Subsequent due dates alone never revoke a completed memory goal; no mastery revocation is introduced here.
`GD-X29`	🟡 PARTIAL	§8.4	An exam goal carries task, reference solution, and grading rubric; valid step points yield \(\Score = \min(\sum_j p_j, \MaxPoints)\), passing at \(\PassingPoints\). The formula legitimizes neither fabricated or negative step points nor ad-hoc task alteration.	M4/M5 (`EXAM`) for authored `examData`; `RT` for grading	Schema, `GVR-016` (tables globally), and `EXAM` (in scopes/floors) validate `examData`; live session grading is not an `M`-gate.
`GD-X30`	🔴 OPEN	§8.4	Normal learning progression checks transitive effective prerequisites. A deliberately chosen direct exam attempt is not blocked by missing preparatory mastery, though a valid task, scope, and authorized session remain required. Direct attempt access itself is not global implicit recognition.	`RT`	Not covered by M7; any future holistic unit-lock profile remains undecided.
`GD-X31`	🔴 OPEN	§8.4	An exam presents the complete unaltered task without substantive hints and grades the full submission against the rubric. A pass is persisted immediately and awaits confirmation; a failed attempt persists neither result nor score and leaves prior mastery intact.	`RT`	Static metadata checks do not prove live session conformance.
`GD-X32`	🔴 OPEN	§8.4	Post-exam feedback reports outcome, score, rationale, and solution, with fresh tasks on retakes. Unlimited retakes remain available; prior exposure to a solution creates no extra mastery tier. Broader implicit-application recognition requires a reviewed local content contract, neither a second personal status flag nor a follow-up exam.	`RT`/`CONCEPT`; independent content review separate	`EXAM`/`CQR-303` do not fully establish the new implicit-recognition contract.
`GD-X33`	⚪ OPT/DOC	§9	A schedule binds target scope, curriculum revision, calendar, capacity, and plan version. Absolute calendar dates never belong on canonical goals, clusters, or placements; rescheduling never alters goal identity.	`OPTIONAL`/`RT`	No curriculum `M`-gate.
`GD-X34`	⚪ OPT/DOC	§9	Time planning is supplementary execution scheduling rather than a fifth personalization level. It prioritizes available goals without creating `requires`, frontier entries, or results; planned acquisition must precede planned downstream use.	`OPTIONAL`/`RT`	Claims no new runtime access validator.
`GD-X35`	⚪ OPT/DOC	§9	Use net available minutes after calendar outages; fit learning activities, reserve buffer, and binding subject budgets without double counting.	`OPTIONAL`	Detailed formulas in `GD-P07`–`GD-P09`; no `M`-gate.
`GD-X36`	🔴 OPEN	§9	Planned target and actual progress share the exact same named scope. Explicit and validly derived implicit completions form total actual progress; manual classroom-coverage markers neither replace nor mask learner completion. Open gaps remain visible, and revisions never rewrite historical events.	`RT`/`CONCEPT`	Joint global actual-progress accounting is not proven by M7.
`GD-X37`	⚪ OPT/DOC	§9	Keep the local planning pilot distinct from the target backend-owned architecture; the learning-unit definition claims no implemented cross-subject budget validator.	`OPTIONAL`/`DOCUMENT`	Local pilot and target architecture must be documented as distinct implementation states.
`GD-X38`	🟡 PARTIAL	§10.1	Official sources justify a goal, first-party visualizations explain it, and supplementary materials support acquisition. None of the three is personal competency evidence.	M1/M2 (`SOURCE`) sources; M7 (`REVIEW`) image QA; `RT` boundary	`SOURCE`, image QA/`REVIEW`, and the content-package validator enforce separated reference types in their scopes; none proves personal competency.
`GD-X39`	🟡 PARTIAL	§10.1	Materials reference the curriculum rather than embedding provider lock-in inside the graph. New optional external materials live in separate content packages bound to stable goal IDs; package selection alters neither scope nor `requires`, mastery, or completion, and material gaps never block learning.	`ARTIFACT`/`RT`	Content-package and link-migration checks enforce separated packages with stable goal IDs in CI; no `M`-gate covers the entire runtime contract.
`GD-X40`	🟡 PARTIAL	§10.1	Sources, mappings, and provenance stay with the curriculum, and first-party visualizations retain their canonical quality contract. Legacy links never establish a permanent second authoring architecture, and returning a URL never proves its content was read.	M1/M2/M7 (`SOURCE`/`REVIEW`) respective evidence; `ARTIFACT`	`GVR-008` forbids legacy goal links globally while `SOURCE` and image QA check evidence in their scopes; coach content fetching is not gated.
`GD-X41`	🟡 PARTIAL	§10.2	Maintain a traceable provenance chain: goal \(\to\) official source \(\to\) concrete section \(\to\) extraction \(\to\) reviewed mapping decision. Justify 1:1, partial, aggregating, or split mappings explicitly rather than claiming verbatim identity everywhere.	M1/M2 (`SOURCE`)	`SOURCE`; technical completeness alone does not substitute for domain source review.
`GD-X42`	🔴 OPEN	§10.2	For `requires` and `contains`, distinguish explicit official-source statements from pedagogical modeling decisions. Never fabricate provenance; keep unresolved, stale, or conflicting evidence visible.	M2 target (`SOURCE`) for provenance; M3 (`ROUTE`) for sequencing review	No current `CQR` gate automatically verifies every individual edge rationale.
`GD-X43`	🟡 PARTIAL	§10.2	Human-readable source rationales are derived views of structured evidence, never a second authoritative register. Keep references and fingerprints consistent across documentation, exports, and runtime; MEM/FWU alignment replaces the official source route only under an explicitly reviewed scope contract.	M1/M2 (`SOURCE`) + `ARTIFACT`/`RT`	Source-rationale and package-conformance checks verify consistency in covered scopes; not a blanket substitute across all interchange formats.
`GD-X44`	🟡 PARTIAL	§10.3	Parser, reference, cycle, projection, route, and content checks carry distinct evidentiary claims and each requires a declared scope. No single migration gate or structural pass certifies them all.	Cross-stage (`BASE`/`ROUTE`/`VIEW`/`REVIEW`)	`BASE`/`ROUTE`/`VIEW`/`REVIEW` are explicitly separated.
`GD-X45`	🟡 PARTIAL	§10.3	Five current content-evidence dimensions: `D` domain-clear description, `P` observable understanding performance, `A` appropriate atomic sizing, `M` justified memory-support decision, and `V` domain-accurate hash-bound visualization.	M5 `A`, M6 `M`, M7 simultaneous `D/P/A/M/V` completion (`REVIEW`)	`REVIEW` enforces `A` (M5), `M` (M6), and the simultaneous `D/P/A/M/V` intersection (M7) without averaging on protected M6/M7 floors; unprotected inventories remain below.
`GD-X46`	🟡 PARTIAL	§10.3	M7 is complete automated curriculum-QA closure atop all lower maturity levels for the declared scope. It implies neither human sign-off, learner trials, nor learning effectiveness; multi-model AI agreement never substitutes for mandatory human decisions.	M7 boundary (`REVIEW`); human evidence outside M0–M7	`CQR-303` enforces automated M7 closure across the five M7 floors; human sign-off, learner trials, and learning effectiveness lie outside M0–M7.
`GD-X47`	🟡 PARTIAL	§10.3	Reviews bind the exact artifact revision; an old image hash never certifies a new image. Content or prerequisite edits can invalidate evidence, technical moves require narrowly justified compatibility, and blanket hash refreshing is forbidden.	M5/M6/M7 per review dimension (`REVIEW`)	`REVIEW` checks current digest bindings; never infer substantive review validity from a mechanical hash update.
`GD-X48`	🟡 PARTIAL	§10.4	Author domain changes in canonical sources and regenerate downstream artifacts; keep published historical archives byte-immutable.	`ARTIFACT`	Plugin-freeze, release, goal-book, and status-diff checks guard canonical regeneration and archive immutability in their CI paths; they grant no publication authorization.
`GD-X49`	🟡 PARTIAL	§10.4	Build JSON and ontology packages from the identical canonical source state; the JSON runtime remains normative. Roundtrips preserve identities, relations, roles, prose, and domain fields per equality contract (semantic equality for texts, byte identity for binary assets) without divergent runtime semantics.	`ARTIFACT`	Package-contract and conformance checks verify JSON/ontology parity in CI (fast mode in routine CI); full standalone roundtrip verification remains a dedicated release check.
`GD-X50`	🟡 PARTIAL	§10.4	Close external goal references completely—including cross-subject prerequisites—without empty placeholders. Complete package closure does not by itself define Personal Curriculum targets.	`BASE` (M0 base for references) + `ARTIFACT`/`VIEW`	`validate:graph`, standalone-builder tests, and the closure compiler check reference closure in their scopes; package closure and Level 2 targets stay separate.
`GD-X51`	🟡 PARTIAL	§10.4	Every verification or release claim names its exact artifact, scope, version, and source digest binding. A local build proves neither host display nor learning effectiveness, and this compendium grants neither release nor production sign-off.	`ARTIFACT`/`DOCUMENT`	Release, submission, and goal-book checkers bind artifact, version, and source hashes in CI; prose claims, live host display, and production approval lie outside.
`GD-X52`	🟡 PARTIAL	§10.5	Function determines license: software, technical schemas, developer docs, and functional coach instructions carry `Apache-2.0`; first-party skill landscapes, content, and didactic media carry `CC-BY-4.0`. There is no blanket dual licensing. Third-party rights stay intact; provenance or AI generation proves neither ownership nor quality; exports carry license, attribution, and modification notices.	`ARTIFACT`	First-party license and package checks enforce `Apache-2.0`/`CC-BY-4.0` declarations on covered repository artifacts in CI; third-party rights clearance is not an `M`-gate.
`GD-X53`	🟡 PARTIAL	§10.5	Private learner results, sessions, and conversations fall neither under the public content license nor into public source or review packages.	`ARTIFACT`/`RT`	Submission-ZIP audits and package-contract checks reject extraneous or private files in CI; end-to-end runtime privacy is not an M7 claim.

U – Recursive Learning Units

Section 11 (03-learning-units). Every row cites its section; GD-* IDs remain stable regardless of later section renumbering.

ID	Status	Section	Rule (concise summary)	Area	Enforcement and boundary
`GD-U01`	🔴 OPEN	§11	A `LearningUnit` recursively binds a learning promise, own atomic goals, and optional child units. Calendar periods are not synonyms; personalization levels, QA maturity, and difficulty retain independent meanings.	`CONCEPT`, no `M`-gate	Not claimed as an implemented runtime feature.
`GD-U02`	🔴 OPEN	§11.1	Unit motivation explains purpose ahead of individual tasks. Only a genuine orientation goal uses the orientation completion contract; a synthetic heading is not a goal, and derived unit completion never claims an orientation dialogue actually took place.	`CONCEPT`/`RT`	Unit display derives no synthetic orientation mastery.
`GD-U03`	🔴 OPEN	§11.1	A pure aggregate unit requires no extra exam. Added integrative synthesis claims may require a synthesis exam, and genuinely new competencies require own acquisition assignment plus valid exam coverage; pure synthesis forces neither artificial ordinary goals nor re-passing child units.	`CONCEPT`; content review independent	`EXAM`/M7 do not prove a complete learning-unit contract.
`GD-U04`	🔴 OPEN	§11.2	\(\Own(U)\) denotes own atomic goals and \(\Children(U)\) mandatory child units after choice resolution. \(T(U) = \Own(U) \cup \bigcup_{V \in \Children(U)} T(V)\) is non-empty. Structural markers are not goals; an aggregate unit may have empty \(\Own(U)\) when its child scope is non-empty.	`CONCEPT`	Future structural contract; no existing `CQR` gate today.
`GD-U05`	🔴 OPEN	§11.2	R4: every acquisition target has exactly one directly owning unit, the smallest authoritative owning unit. Viewing a parent never turns child goals into parent-owned goals; overlapping references unite as sets, and focus switches alter neither ownership nor local recognition contracts.	`BASE` (validity condition of every authored unit definition, same class as acyclicity): hard from the first authored unit; a violating unit cannot be merged	No unit definitions and no unit validator exist on `main`; the placement requires a fail-closed check from the first authored unit. Multi-view display creates no duplicate acquisition or progress counter.
`GD-U06`	🔴 OPEN	§11.2	R1: every own ordinary content goal lies on a prerequisite path from at least one own assessment of the unit, directly or transitively; a path that leads only to a memory goal does not count. The reviewed application coverage of such an assessment is R2 (`GD-A08`). Without both the local contract is invalid; arbitrary manual waiver masks never substitute for it.	`BASE` (validity condition of every authored unit definition, same class as acyclicity): hard from the first authored unit; a violating unit cannot be merged; content share R2 see `GD-A08`	No unit definitions and no unit validator exist on `main`; `ROUTE`/`EXAM` check related properties rather than this unit condition; no fabricated M7 gate.
`GD-U07`	🔴 OPEN	§11.2	Authoritative program/placement or reviewed composition structure defines a learning unit. Expanding an arbitrary `contains` cluster never turns it into a learning unit; general cluster satisfaction stays distinct.	`CONCEPT`/`VIEW`	Clusters are never automatically classified as learning units at runtime.
`GD-U08`	🔴 OPEN	§11.3	\(\Complete(U)\) holds iff all own goals lie in \(\Meff\) and all mandatory child units are complete. Neither extra unit-mastery state is stored nor is mastery written backward onto children.	`CONCEPT`	Document-model verification is neither a runtime nor an `M`-gate.
`GD-U09`	🔴 OPEN	§11.3	Exam and memory goals remain explicit; passing a parent exam replaces neither a child exam nor verified recall. Synthesis assessment is an additional own obligation, card decks alone never prove application, and child recall obligations must never be hoisted onto parents to fake child completion.	`CONCEPT`/`RT`	Never infer runtime fulfillment from M6 card review or M7 content status.
`GD-U10`	🔴 OPEN	§11.4	Own ordinary competencies can be implicitly recognized ahead of independent memory, child, or parent completion. Recognize only own goals of the local contract—never via a second parent contract spanning all descendants. Open parent synthesis never revokes valid child recognition.	`CONCEPT`	Global runtime effect is not yet implemented; see `GD-A10`–`GD-A14`.
`GD-U11`	🔴 OPEN	§11.5	An explicitly authorized whole-unit prerequisite denotes full unit completion and lives outside \(C\) and \(R\). An atomic prerequisite requires only that competency rather than the entire foreign unit; neither compiler nor coach may conflate the two.	`CONCEPT`; access policy `OPTIONAL`	An atomic edge induces no whole-unit hard lock.
`GD-U12`	🔴 OPEN	§11.5	The unit hierarchy is acyclic. Any claimed undivided unit-completion sequence must additionally be checked for block cycles; an atomic DAG alone does not suffice. Cyclic block grouping requires regrouping or finer interleaved sequencing rather than inducing an implicit unit-wide lock.	`CONCEPT` / conditional structural contract	Finite document model verifies examples; no production unit validator is claimed.
`GD-U13`	🟡 PARTIAL	§11.5	Transitive minimality remains binding for authored `requires` in accordance with the atomic target norm and ordered migration. No extra minimality rule is posited for a not-yet-decided holistic unit-access relation.	`BASE` (M0 target base before M1 per `GD-C11`/`GD-C16`); unit relation `OPTIONAL`	The authored-`requires` share is enforced as in `GD-C11`/`GD-C16` (`CQR-001`, hard via protected floors, canonical Gymnasium catalog only); no unit relation exists to check. This shared minimality reference never authorizes deleting atomic edges in favor of remaining inherited cluster requirements.
`GD-U14`	⚪ OPT/DOC	§11.5	Both a hard unit-access profile and direct exam access under it remain undecided. The illustrative equations \(\operatorname{Open}(U) = \bigwedge_{V \in \operatorname{Pred}(U)} \operatorname{Cleared}(V)\) and \(\operatorname{Cleared} = \operatorname{Open} \land \Complete\) apply only if an access profile is later authorized. \(\Complete\) denotes evidentiary state and \(\operatorname{Cleared}\) course policy; neither grants mastery, and the formula does not settle direct-exam exceptions.	`OPTIONAL`, no curriculum `M`-gate	Claim neither authorization nor implementation of a technical unit lock.
`GD-U15`	🔴 OPEN	§11; §11.1	A unit establishes secure independent application of its promise before planned learning builds on it; its assessments expose essential misconceptions, and a gap leads to focused learning and a fresh equivalent assessment. Boundaries follow meaningful conceptual transitions, not existing assessments, target quotas, headings, or edge density; edges are never changed to fit a cut. The ordinary grading contract applies: no hidden pass rule, no second stored outcome.	`CONCEPT`; content review (`REVIEW`) separate; no access lock	Didactic design and review criterion; no validator or `M`-gate checks unit boundaries or purpose, and no unit is authored on `main`.

P – Optional Acquisition and Time Planning

Section 12 (04-planning). Every row cites its section; GD-* IDs remain stable regardless of later section renumbering.

ID	Status	Section	Rule (concise summary)	Area	Enforcement and boundary
`GD-P01`	⚪ OPT/DOC	§12	An optional schedule assigns acquisition targets to calendar levels; it neither defines learning units, alters core skill-graph validity or mastery, nor introduces holistic unit locks or blanket full-year completion gates.	`OPTIONAL`	Claims no new production budget validator.
`GD-P02`	⚪ OPT/DOC	§12.1	A plan specifies atomic acquisition targets \(T\), a disjoint declared entry basis \(B\), a positive number of levels, and a unique first-acquisition level \(\ell\) per target defining level sets and prefixes. Later practice or application is scheduled activity rather than a duplicate goal ID. \(B\) writes no learner mastery; personal entry gaps require diagnosis and time.	`OPTIONAL`	No curriculum `M`-gate.
`GD-P03`	⚪ OPT/DOC	§12.2	The hypothetical plan relation prunes outgoing requirements solely at the declared entry basis \(B\), preserving full transitive closure across all planned levels. This planning convention never permits truncating real runtime prerequisite traversal at mastered intermediates.	`OPTIONAL`	Purely hypothetical planning construct, not a runtime mastery contract.
`GD-P04`	⚪ OPT/DOC	§12.2	A completion witness orders every level's targets so each enters the global frontier given \(B\), prior levels, and preceding witness atoms. Inherited, cluster, and cross-subject requirements remain intact, with neither normal-mode hiding, direct exam jumps, nor anticipated implicit recognition.	`OPTIONAL`	No curriculum `M`-gate; not proven by current UI navigation paths.
`GD-P05`	⚪ OPT/DOC	§12.2	Required atoms equal the full closure under the plan relation. Necessarily, every target's prerequisite demand lies in \(B\) or at latest its own level: outside \(B \cup T\) is unplanned, a higher level is an ordering error, and same-level dependencies are permitted. Graph and expansion validity precede closure checking.	`OPTIONAL`	Structural feasibility check only; does not by itself prove time feasibility.
`GD-P06`	⚪ OPT/DOC	§12.2	The finite fixed-point iteration starts from \(B\) plus prior levels and repeatedly adds currently frontier-eligible level targets, succeeding iff the entire level is reached. Eligibility is monotone, each parallel wave is serializable, the test mutates no learner data, and a purely atomic DAG reduces to standard topological sorting.	`OPTIONAL`/`DOCUMENT`	Algorithmic specification, not claimed as a production validator.
`GD-P07`	⚪ OPT/DOC	§12.3	Any time-feasibility claim states reference learners, entry state, target/elective selection, calendar, estimation method, and uncertainty in a shared calibrated time unit: \(\Work \geq 0\), \(\Reserve \geq 0\), \(\Budget > 0\), and \(\Work + \Reserve \leq \Budget\). Keep reserve protected (or disclose zero explicitly) without double-counting exam or review effort.	`OPTIONAL`	Without calibrated estimates there is no time-validity claim; no `M`-gate.
`GD-P08`	⚪ OPT/DOC	§12.3	The activity model counts actual effort once across acquisition, practice, recall, assessment, and supplementary work. A shared canonical atom is not acquired twice across subjects, though subject-specific application may still consume time. `weight` and uncalibrated goal counts are never minutes.	`OPTIONAL`	Underlying estimation model and empirical data must be evidenced separately.
`GD-P09`	⚪ OPT/DOC	§12.3	Worked numerical examples are not empirical curriculum durations. Binding subject, room, or date sub-budgets must hold simultaneously without double-counting shared time, and concrete prerequisites must be scheduled ahead of their planned application.	`OPTIONAL`	Satisfying a total budget alone is insufficient when sub-constraints apply.
`GD-P10`	⚪ OPT/DOC	§12.4	Full validity of an optional plan requires both structural and time feasibility under bound assumptions. Without calibrated time data only structural feasibility may be claimed; every claim binds plan, revision, scope, and estimation assumptions without guaranteeing uniform individual learning speed.	`OPTIONAL`	Neither an M7 gate nor proof of learning effectiveness.
`GD-P11`	⚪ OPT/DOC	§12.4	Unit completion follows own and child obligations; learning units and calendar periods may cross-cut one another. A schedule profile never creates a blanket full-year prerequisite for later goals.	`OPTIONAL`/`CONCEPT`	Access policy remains separate.

A – Implicit Mastery via Local Assessment Contracts

Section 13 (05-assessment). Every row cites its section; GD-* IDs remain stable regardless of later section renumbering.

ID	Status	Section	Rule (concise summary)	Area	Enforcement and boundary
`GD-A01`	🔴 OPEN	§13	Explicit ordinary completion evidences understanding and a valid local exam evidences independent application; both act globally. Exactly one binary state is stored—without backward writes, extra qualification tiers, manual waiver profiles, or second sub-exams. Exam and memory goals themselves always remain explicit.	`CONCEPT`, no `M`-gate	New semantics is explicitly not yet implemented; never reinterpret legacy tests as proving it.
`GD-A02`	🔴 OPEN	§13.1	Each local contract binds authoritative own unit goals, canonical IDs, curriculum revision, and the reviewed evidentiary scope of own exams. Total child scope never forms an inflated parent recognition contract, and display, focus, or schedule roots never alter unit ownership.	`CONCEPT`	Future contract and revision check; neither `CQR-401` nor M7 substitutes for it.
`GD-A03`	🔴 OPEN	§13.1	Determine exam and memory goals authoritatively; \(P_U = Q_U \cup A_{\mathrm{mem},U}\) protects all own goals of both types from implicit completion. Orientation is not a protected demand root, though a derived unit marker claims neither competency nor dialogue. External support neither inflates the unit denominator nor permits local recognition of foreign \(\Own\) goals.	`CONCEPT`	Memory content review alone does not establish this recognition logic.
`GD-A04`	🔴 OPEN	§13.1	Operate on the fully inherited and atomically expanded relation—including hidden and cross-subject prerequisite support—with all references resolvable and the normalized graph acyclic. Compute full closure prior to display filtering.	`BASE` (M0 target base for inputs); `CONCEPT` for recognition use	`BASE` alone does not prove the recognition normalizer.
`GD-A05`	🔴 OPEN	§13.2	\(N_U = \operatorname{Reach}_{\Rat}^{*}(P_U \setminus M)\) comprises all open protected roots and every transitive prerequisite without stopping at mastered intermediates. \(I_U = \Own(U) \setminus (M \cup P_U \cup N_U)\) is disjoint from \(M\) and \(P_U\). Redundant edges with identical closure leave \(I_U\) unchanged.	`CONCEPT`; formula check `DOCUMENT`	Legacy contextual waiver logic does not implement or prove this equation.
`GD-A06`	🔴 OPEN	§13.2	The demand traversal visits reachable atoms \(V_U^{\mathrm{reach}}\) and their outgoing \(\Rat\) edges \(E_U^{\mathrm{reach}}\) at most once; computing \(I_U\) adds one pass over \(\Own(U)\), yielding \(O(\lvert\Own(U)\rvert + \lvert V_U^{\mathrm{reach}}\rvert + \lvert E_U^{\mathrm{reach}}\rvert)\) without materializing a quadratic closure.	`CONCEPT` / algorithmic contract	Verified in the finite document model, not a production benchmark.
`GD-A07`	🔴 OPEN	§13.3	The implicit-set equation is evaluated only for units whose contract satisfies R1–R4 (`GD-U06`, `GD-A08`, `GD-A18`, `GD-U05`) with reviewed coverage. Otherwise the unit's own ordinary goals stay open and the violation keeps the curriculum from reaching the respective milestone; a goal that no assessment covers stays open and is a rule violation, never recognized for free. Absence of a downstream consumer proves nothing, and arbitrary waiver masks are forbidden.	`CONCEPT` for the recognition gate; violations count against `BASE` (R1, R3, R4) and M5/M7 (R2)	Recognition is not implemented and no unit validator exists on `main`; `ROUTE` is structurally related but not equivalent to a local unit contract.
`GD-A08`	🔴 OPEN	§13.3	R2: for every own ordinary goal, at least one own assessment that requires it has reviewed evidence that its tasks and normal grading really test the goal, so that no essential skill can be fully compensated away by unrelated rubric points. Neither `requires` paths, titles, `coveredGoalIds`, passing thresholds, nor DAG validity prove this; independent review must verify all tasks, solutions, rubrics, and every passing score combination rather than just their union.	M5 (`EXAM`/`REVIEW`: assessment approval and coverage metadata), current per-goal binding M7 (`REVIEW`); without it the curriculum does not reach the milestone	`CQR-203` checks approval status and coverage metadata of terminal assessments in configured scopes, and `CQR-303` current per-goal `D/P/A/M/V` evidence; neither reviews unit-local coverage, and none of the six required M7 checks is a coverage check. Not enforced today.
`GD-A09`	🔴 OPEN	§13.3	Passing an exam remains a single binary learner outcome. Its broader evidentiary reach resides in the pre-reviewed exam contract rather than a second personal status flag; legacy unreviewed exams are never retroactively upgraded by diagrams or software.	`CONCEPT`/`RT`	Stronger evidentiary meaning resides in reviewed exam content; no second stored learner status follows.
`GD-A10`	🔴 OPEN	§13.4	The valid contract set \(\mathcal U_{\mathrm{valid}}\) is bound durably per learner and revision—neither restricted to the active view/plan nor expanded across all latest revisions. \(\Meff = M \cup \bigcup_{U \in \mathcal U_{\mathrm{valid}}} I_U\) stays available across focus or scope changes and derives strictly from explicit results \(M\) without feeding prior implicit sets back into demand.	`RT`/`CONCEPT`: versioned contract binding (R5, `GD-A12`) is a precondition before implicit recognition is enabled in production	Not yet claimed in production; no curriculum `M`-gate.
`GD-A11`	🔴 OPEN	§13.4	Subsequent exams or units never reopen an already recognized goal. Under fixed contracts, additional explicit mastery can only grow \(\Meff\), and moving a goal from \(I_U\) into \(M\) preserves satisfaction monotonically.	`CONCEPT`; monotonicity proof `DOCUMENT`	Finite model checking \(\neq\) production runtime conformance.
`GD-A12`	🔴 OPEN	§13.4	R5: a later re-cut of units or later edges never grant additional recognition to existing results; the contract is versioned. A stable exam ID carries stable recognition semantics, and task variants satisfy the same contract. Changed claims require a new assessment identity or reviewed migration; unverified historical exams retain explicit pass status without fabricating broader implicit application. Separate genuine error correction from mere plan or consumer changes.	`RT` (runtime contract): precondition before implicit recognition is enabled in production; `ARTIFACT` for contract versions	Not implemented; no recognition runtime exists yet to enable. No extra personal qualification record; no silent historical upgrade.
`GD-A13`	🔴 OPEN	§13.5	Full unit completion is recursive—equivalent under complete goal representation to \(T(U) \subseteq \Meff\), with own \(P_U \subseteq M\). Passing own exam and memory goals empties \(N_U\) and can recognize own ordinary goals, never child units. Open parent synthesis never revokes child recognition, and passing a parent exam never replaces child exam or memory goals.	`CONCEPT`	Specification equation, not an existing M7 feature.
`GD-A14`	🔴 OPEN	§13.5	An open own memory goal blocks full unit completion, yet withholds implicit recognition from another own goal only if it transitively requires that goal. A shared foundation stays demanded as long as any local open protected consumer requires it; a later open consumer in another unit contract never revokes existing recognition.	`CONCEPT`	Never conflate local contract demand with downstream units.
`GD-A15`	🔴 OPEN	§13.5	Direct exam attempt access remains the baseline given a valid task, auth, session, and explicit selection. Optional unit-wide locking stays undecided; the learning-unit concept creates neither a lock nor an auth bypass. Normal learning progression evaluates \(\Meff\) across full effective transitivity under separate normal/strict focus rules.	`RT`/`CONCEPT`/`OPTIONAL`	Neither an `M`-gate nor an access-lock policy decision is introduced here.
`GD-A16`	🔴 OPEN	§13.6	Cumulative actual progress counts \(\lvert T_{\mathrm{Plan}} \cap \Meff \rvert\), uniting explicit, implicit, and multi-contract contributions prior to counting. Any explicit-only view must state its distinct numerator/denominator. Current-period reporting counts genuinely dated explicit completion events rather than newly derived implicit items, prior work, or later scope expansion.	`RT`/`CONCEPT`	M7 checks no personal plan or period accounting.
`GD-A17`	🔴 OPEN	§13.6	Static contract structures may be cached per revision, whereas derived implicit completions never form a second persisted source of truth. Invalid or unverified contracts stay visible as explicit limitations. Automated formula checks replace neither independent content review nor human trials and learning-effectiveness evidence.	`CONCEPT`/`DOCUMENT`; evidentiary boundary cross-stage	No curriculum `M`-gate covers the full persistence and disclosure contract.
`GD-A18`	🔴 OPEN	§13.3	R3: an assessment claims coverage only for goals it requires directly or transitively: \(\operatorname{Cov}(q) \subseteq \{b \in A \mid (q,b) \in \Rat^+\}\). This guarantees that the assessment which actually tests a goal is among the open protected goals the recognition rule waits for. A missing path is a subject-review finding, never a reason for a redundant direct edge.	`BASE` (validity condition of every authored unit definition, same class as acyclicity): hard from the first authored unit; a violating unit cannot be merged	No unit definitions and no unit validator exist on `main`. Related: `CQR-104` binds every `coveredGoalIds` entry to a visible prerequisite path from the assessment, but only for terminal assessments of canonical Economics; `CQR-203` only checks that coverage metadata exists.

D – Document Appendices: Status, Verification Strategy, Sources, and Sign-Off

Appendices B–F (06-appendices). Every row cites its appendix; GD-* IDs remain stable regardless of later section renumbering.

ID	Status	Section	Rule (concise summary)	Area	Enforcement and boundary
`GD-D01`	⚪ OPT/DOC	Appendix B	Bind every implementation-status statement to a named audited commit. Historical commit `616324e857` and build manifests make no claim about live `HEAD`. A document build proves neither runtime conformance nor deployment, and global implicit recognition remains marked unimplemented.	`DOCUMENT`	Dated implementation audit findings are not a live conformance check.
`GD-D02`	⚪ OPT/DOC	Appendix B	List historical implementation divergences individually: completion traversal truncation, new protected set and local coverage requirements, contextual waivers vs. durable global contracts, incomplete validity proofs, and `core`-preferring cluster aggregation vs. all atoms. Never present mere test existence as remediation.	`DOCUMENT`/`CONCEPT`	Dated audit findings, not automatically a live fail/pass list.
`GD-D03`	⚪ OPT/DOC	Appendix B	The memory goal type carries its own explicit completion record; review due dates never revoke completed mastery, and ordinary completed goals receive no review due dates. Learning units remain an unimplemented concept; unit-wide locks and direct exam access under them stay open, and time planning stays separate.	`DOCUMENT`/`CONCEPT`/`OPTIONAL`	Keep the binding memory-goal decision distinct from dated implementation findings; explicit card-goal completion, SRS scheduling, and undecided unit-access policy are separate contracts.
`GD-D04`	⚪ OPT/DOC	Appendix C	The mathematical model checker is a reference verification script that reads no learner data and mutates no runtime state. Exhaustive finite DAG, mastery, and protected-set checks hold under assumed complete local assessment coverage; the cited 5,222 cases are a dated document finding.	`DOCUMENT`	Neither a curriculum `M`-gate nor end-to-end production verification.
`GD-D05`	⚪ OPT/DOC	Appendix C	Reduction/closure checks and inheritance counterexamples prove their exact algebraic claims; testing raw edges against inheritance does not prove the new atomic migration priority. Reports must keep the binding target norm distinct from the bounded model test.	`DOCUMENT`; `G17` target base (`BASE`) before M1 stays separate, while M3 (`ROUTE`) covers additional routes	The bounded model test proves its stated algebraic claims, not an implemented gate for the new atomic migration order.
`GD-D06`	⚪ OPT/DOC	Appendix C	Monotonicity, protection of exam and memory goals, recursive unit cases, block-cycle detection, downstream contracts/external support, and error cases are distinct mathematical checks. The 128/948 finite test cases do not constitute full curriculum or runtime conformance, and block-cycle checking creates no lock policy.	`DOCUMENT`	Historical model counts stand strictly as dated figures unless a fresh run is executed.
`GD-D07`	⚪ OPT/DOC	Appendix C	Diagram verification checks displayed sets and, under `--check`, deterministic PNG bytes. It proves neither task quality nor grading reliability, identity verification, or learning effectiveness; independent content review and runtime, security, concurrency, and UI tests remain separate.	`DOCUMENT`	Proves neither M7, live host rendering, nor human-trial outcomes.
`GD-D08`	⚪ OPT/DOC	Appendix D	Keep symbol notation consistent throughout: \(K\) denotes structural clusters, \(A_{\mathrm{mem}}\) and \(A_{\mathrm{mem},U}\) memory goals, \(\operatorname{Cards}(g)\) their flashcards, \(\Rat\) the atomic relation, \(M/\Meff\) explicit/effective mastery, and \(\Own/T\) direct/total unit scope. \(m=0\) does not preclude implicit recognition, and glossary definitions introduce no extra outcome tiers.	`DOCUMENT`	Notation consistency check across `GD-C`/`U`/`A`, introducing no extra substantive rule.
`GD-D09`	⚪ OPT/DOC	Appendix E	Chapter source attributions record normative ownership; cross-referencing a concept never claims production implementation of optional profiles. Never silently import norms from unread external files into this register.	`DOCUMENT`	Source attribution register, not an `M`-gate.
`GD-D10`	⚪ OPT/DOC	Appendix E	Attribute static original illustrations and generated deterministic PNGs accurately; equations and legends remain normative while figures create no extra edges. Technical prose and build tooling carry `Apache-2.0`, first-party didactic figures `CC-BY-4.0` where rights hold, with clear provenance and revision metadata.	`DOCUMENT`/`ARTIFACT`	Figure byte and license boundaries are independent of curriculum M7.
`GD-D11`	⚪ OPT/DOC	Appendix F	The build pipeline binds chapters and PNG assets and documents prerequisites and exact commands. Multi-pass LuaLaTeX resolves cross-references, and the manifest records SHA-256 digests of inputs and PDFs alongside dates and tool versions; differing TeX distributions do not guarantee byte-identical PDFs.	`DOCUMENT`	Document-build conformance, neither a runtime nor a content pass.
`GD-D12`	⚪ OPT/DOC	Appendix F	Document sign-off requires an error-free LaTeX build, resolved cross-references, no unchecked overfull horizontal or vertical boxes, and visual PDF inspection. Mathematical model checking complements sign-off without replacing runtime, domain, or empirical validation.	`DOCUMENT`	Document sign-off criteria outside curriculum M0–M7.

Completeness Summary

Part	Identifiers	Count
C – Formal Core	`GD-C01`–`GD-C36`	36
X – Context, Views, Goal Types, Provenance, and Quality	`GD-X01`–`GD-X53`	53
U – Recursive Learning Units	`GD-U01`–`GD-U15`	15
P – Optional Acquisition and Time Planning	`GD-P01`–`GD-P11`	11
A – Implicit Mastery via Local Assessment Contracts	`GD-A01`–`GD-A18`	18
D – Document Appendices: Status, Verification Strategy, Sources, and Sign-Off	`GD-D01`–`GD-D12`	12

Completeness and Evidentiary Boundary. This register comprises 36 core (C), 53 context (X), 15 learning-unit (U), 11 planning (P), 18 implicit-recognition (A), and 12 document (D) rule groups, totaling 145 rule groups. Distinct sub-obligations are preserved within each row while repeated explanations and examples map onto those same obligations. Target assignment claims neither complete enforcement nor a live CI, runtime, release, or learning-effectiveness pass.

Sources: Status generator (rule catalog, deriveScopeMaturity, deriveCurriculumMaturity); skill-graph validator; transitive-minimality auditor; graph-integrity adapter; prerequisite-reduction proof registry; floor checker; floor policy; composition validator; filter validator; schema validator; UUID validator; content-package validator; package-contract check; first-party license check; command registry; CI workflow; maturity and route documentation; graph validation rules; Markdown matrix.