The schema
The JSON Schema contracts that every sidecar must satisfy
Each authored sidecar is validated against its JSON Schema before any derivation. A field marked required must be present; fields with an enum accept only the listed values. These schemas are the executable Tier A contract — this page renders them directly, so it never drifts from what the validator enforces. v2 (ADR-009) is canonical; v1 remains accepted during the migration window.
One file describes exactly one (jurisdiction x assessment_system x year). v2 (ADR-008/ADR-009) adds: the enrollment-grade model on every content area (intended_enrollment_grade fixed|variable + enrolled_grades_tested — disentangling the instrument's target grade, the student's enrolled grade, and cut keys), scale_bounds (loss/hoss keyed by enrolled grade, mirroring cutscores), a canonical assessment_type enum with conditional measurement.elp / measurement.alternate extension blocks, source_documents[], and per-grade cutscores_source confidence. AXIS RULE: cutscores, scale_bounds, and cutscores_source are ALWAYS keyed by ENROLLED grade — never by instrument or form name; instrument forms (e.g. WIDA grade clusters) are measurement facts in the extension blocks. v1 records remain valid during the migration window (ADR-009 D6).
Assessment System Metadata (self-contained per-annual sidecar) v2 · https://centerforassessment.github.io/assessment-metadata-registry/schemas/amr.assessment.v2.json
Canonical v2 identifier. Supersedes amr.assessment_system.v1 (and the sgpc.assessment_metadata.v0.1 alias), which remain accepted by the validator during the migration window.
Governance status. draft = unreviewed (incl. machine-drafted or placeholder values); reviewed = a human checked it; verified = confirmed against an authoritative citation; deprecated = retained for history but superseded. Colleague verification-state map: unverified->draft, auto_derived->draft(+low), in_review->reviewed, human_verified->verified.
draft reviewed verified deprecatedConfidence in the substantive facts (esp. cutscores/levels/bounds). Placeholder/scaffold values are low. Per-value confidence lives on scale_bounds entries (source) and cutscores_source.
low medium highWhere this record came from, who entered it, and what changed. Required (with source_citation) for any status other than draft.
URL or document reference for the authoritative source of the substantive claims (the primary citation; additional evidence goes in source_documents).
Author identifier; for machine-drafted records use the form ai:<harness-id>.
Who performed the last verification (reviewer identifier), or null if unverified. Pairs with last_verified_at (ADR-010, adopted from the colleague spec's provenance).
What changed vs the prior year/commit and why (assessment-system change vs metadata correction vs modeling decision).
Evidence list beyond the primary provenance.source_citation: technical manuals, interpretive guides, state memos. url may be null when a document is known but not yet located online (ADR-009 D4).
state territory consortium district otherCanonical discriminator (ADR-008 SS5). Legacy v1 strings (state-summative, english-language-proficiency, general) are normalized by migrate_registry(). elp permits measurement.elp; alternate permits measurement.alternate.
summative alternate elp science end-of-courseFree-text note about the provenance/status of the cutscores/scale_bounds blocks specifically (e.g. a PLACEHOLDER warning). Complements the structured provenance block and per-grade cutscores_source.
Year-resolved comparability of this administration to the prior year: scale breaks, vendor/blueprint changes, and administration gaps (e.g. COVID-19).
The enrollment-grade model (ADR-009 D1): the relationship between this instrument and the ENROLLED grade of the student taking it. fixed = the instrument targets one enrolled grade (ILEARN Grade 8 Math); variable = students from multiple enrolled grades sit the same instrument (ACT; WIDA grade-span forms).
fixed variableOrdered enrolled grades that sit this content area. K (and PK) permitted. EOC without a grade association may use [] with an explanatory note.
Free text for the murky edge cases (e.g. 'ACT is a high-school test usually administered to 11th graders but taken by 10th/12th for graduation-competency purposes').
Canonical proficiency benchmark (ADR-010): the label (one of labels[]) of the LOWEST level counted as proficient. A single value that is robust to label edits and cleanly separates the measurement fact (the levels exist) from the policy fact (where proficiency begins). Consumers derive the boolean mask from it. Preferred over the legacy proficient[] mask; the validator checks it names a real label.
DEPRECATED (ADR-010): legacy boolean mask aligned positionally to labels[]. Superseded by proficient_from (a single label), which avoids positional coupling and the non-monotonic masks the array form permits. Still accepted during migration; when both are present the validator checks they agree.
content_area -> ENROLLED grade -> strictly increasing level lower bounds (count = n_levels - 1). Grade keys are enrolled grades, never instrument/form names (axis rule). EXCEPTION (ADR-010): an end-of-course assessment, whose standard is instrument-level rather than grade-specific, may use the single sentinel key "eoc" in place of enrolled grades — write the cut once instead of copying it across every enrolled grade. Permitted by the validator only when assessment_type = end-of-course.
Optional per-grade confidence for cutscore vectors: content_area -> enrolled grade -> official|derived|provisional (ADR-008 SS5 per-value confidence).
content_area -> ENROLLED grade -> scale envelope {loss, hoss, source?}. Mirrors cutscores keying exactly (ADR-009 D2), including the end-of-course "eoc" sentinel key (ADR-010); per-grade because scales may differ by grade (ILEARN); uniform scales simply repeat values. Invariant (validator): loss <= min(cuts) <= max(cuts) <= hoss where both blocks are present.
Lowest obtainable scale score.
Highest obtainable scale score.
official derived provisionalType-conditional measurement extension blocks (ADR-009 D3). Vendor/psychometric facts ONLY — policy facts (ELP exit, growth targets, timelines, participation rules) live in the accountability record (ADR-002 heuristic).
English-language-proficiency measurement facts (permitted only when assessment_type = elp).
Composite score definitions keyed by composite id (which may match a content_areas[].id, e.g. ELP_COMPOSITE): label + domain weights.
Instrument FORM spans (e.g. K, 1, 2-3, 4-5, 6-8, 9-12). These are forms, not score-interpretation keys — cuts and bounds are keyed by enrolled grade (axis rule).
Alternate-assessment measurement facts (permitted only when assessment_type = alternate). Participation criteria and the federal 1% cap are POLICY facts -> accountability record.
scale profileCrosswalks: canonical id -> alternate names/codes (old names, SGP content-area names, colleague 'subjects' names, state abbreviations).
Optional Ed-Fi descriptor crosswalk overrides for this cell.
Optional free-text interpretive note for this administration (how to read the metadata; policy caveats). Distinct from provenance.source_citation and cutscores_provenance (source/citation). A note may also attach per content_area, achievement_levels entry, and scale_bounds cell.
One file describes exactly one (jurisdiction x accountability_system x year). The accountability system is a state-wide object that USES one or more assessment systems: it carries the policy goals a score is measured against, year-resolved, each cross-linking an assessment system + content area (ADR-002). v2 (ADR-009 D5) adds the policy blocks reclassified out of the colleague spec's measurement extensions: growth_targets, timelines, and participation, plus source_documents[]. targets[] is unchanged from v1 — the SGPc re-merge contract is untouched. growth_targets shape is provisional and hardens when the first real record is authored.
Accountability System Metadata (self-contained per-annual sidecar) v2 · https://centerforassessment.github.io/assessment-metadata-registry/schemas/amr.accountability.v2.json
Canonical v2 identifier. Supersedes amr.accountability_system.v1, which remains accepted by the validator during the migration window.
Governance status; non-draft requires provenance.source_citation.
draft reviewed verified deprecatedlow medium highEvidence list beyond the primary provenance.source_citation (state board rules, ESSA plans, exit-criteria memos). url may be null when known but not yet located (ADR-009 D4).
state territory consortium district othere.g. ESSA, state-specific. Optional.
Criterion-referenced achievement targets: the policy goal a score is measured against. Each target cross-links the assessment system + content area it applies to. Unchanged from v1 (SGPc re-merge contract).
Cross-link to an assessment record (same jurisdiction + year).
exit (ELP reclassification) or proficiency (summative accountability goal).
exit proficiencyproficiency_boundary resolves against the linked assessment's cutscores + proficient mask; scale_score carries explicit per-grade thresholds; level is reserved.
proficiency_boundary scale_score level>= >Required for basis='scale_score': ENROLLED grade -> scale-score threshold (axis rule: grade keys are enrolled grades).
For basis='level' (deferred): the fractional level target (e.g. 4.2).
Expected growth toward exit/proficiency (reclassified from colleague elp$growth_targets — ADR-009 D5). PROVISIONAL SHAPE: required cross-link + description; per-grade values optional. Harden when the first real record is authored.
The growth expectation rule in prose (e.g. 'gain >= 0.5 proficiency levels per year toward exit').
Optional ENROLLED grade -> numeric growth expectation.
Exit/on-time policy timelines (reclassified from colleague elp$timelines — ADR-009 D5).
Long-term English learner definition (e.g. '> 6 years without exit').
Participation policy (reclassified from colleague alternate$participation_criteria / federal_cap — ADR-009 D5).
Participation criteria statements (e.g. alternate-assessment eligibility rules).
Federal alternate-assessment participation cap as a proportion (e.g. 0.01).
Optional free-text interpretive note for this accountability administration. Distinct from provenance.source_citation; a note may also attach per target.
One file describes exactly one (jurisdiction x assessment_system x year). Identity (jurisdiction + assessment_system) is invariant and conflict-checked at ingest; administration, assessment_program, content_areas, achievement_levels, and cutscores are year-resolved (the file is the year), so performance-level re-establishment is first-class. Ported from SGPc sgpc.assessment_metadata.v0.1; the legacy schema_version string is accepted as an alias during migration.
Assessment System Metadata (self-contained per-annual sidecar) v1 · https://centerforassessment.github.io/assessment-metadata-registry/schemas/amr.assessment_system.v1.json
amr.assessment_system.v1 is canonical; sgpc.assessment_metadata.v0.1 is the accepted legacy alias during migration.
amr.assessment_system.v1 sgpc.assessment_metadata.v0.1Governance status. draft = unreviewed (incl. machine-drafted or placeholder values); reviewed = a human checked it; verified = confirmed against an authoritative citation; deprecated = retained for history but superseded.
draft reviewed verified deprecatedConfidence in the substantive facts (esp. cutscores/levels). Placeholder/scaffold values are low.
low medium highWhere this record came from, who entered it, and what changed. Required (with source_citation) for any status other than draft.
URL or document reference for the authoritative source of the substantive claims.
Author identifier; for machine-drafted records use the form ai:<harness-id>.
What changed vs the prior year/commit and why (assessment-system change vs metadata correction vs modeling decision).
state territory consortium district othere.g. state-summative, english-language-proficiency, alternate, science, end-of-course.
Free-text note about the provenance/status of the cutscores block specifically (e.g. a PLACEHOLDER warning). Retained from the SGPc sidecars; complements the structured provenance block.
Year-resolved comparability of this administration to the prior year: scale breaks, vendor/blueprint changes, and administration gaps (e.g. COVID-19). Lets consumers flag non-comparable growth spans and the changelog surface breaks.
Are scores comparable to the immediately prior administered year? null = unknown.
Did the scale change in this year (a break vs the prior year)?
Name of the superseded scale when scale_transition is true.
Was this administration given at all? false records an intentional gap (e.g. 2020 COVID) as a first-class fact.
Crosswalks: canonical id -> alternate names/codes (old names, SGP content-area names, state abbreviations).
Optional Ed-Fi descriptor crosswalk overrides for this cell. Each map overrides the bundled descriptor-profile defaults; absent keys fall back to the profile.
One file describes exactly one (jurisdiction x accountability_system x year). The accountability system is a state-wide object that USES one or more assessment systems: it carries the policy goals a score is measured against (ELP exit thresholds, summative accountability proficiency targets), year-resolved. Each target CROSS-LINKS an assessment system + content area rather than living inside the assessment sidecar, because the target is state policy (see ADR-002), not a property of the assessment. Extensible: participation rules, N-size, indicator weights, and business rules are future additions.
Accountability System Metadata (self-contained per-annual sidecar) v1 · https://centerforassessment.github.io/assessment-metadata-registry/schemas/amr.accountability_system.v1.json
Governance status; non-draft requires provenance.source_citation.
draft reviewed verified deprecatedlow medium highstate territory consortium district othere.g. ESSA, state-specific. Optional.
Criterion-referenced achievement targets: the policy goal a score is measured against. Each target cross-links the assessment system + content area it applies to.
Cross-link to an amr.assessment_system.v1 record (same jurisdiction + year).
exit (ELP reclassification) or proficiency (summative accountability goal).
exit proficiencyproficiency_boundary resolves against the linked assessment's cutscores + proficient mask; scale_score carries explicit per-grade thresholds; level is reserved.
proficiency_boundary scale_score level>= >Required for basis='scale_score': grade -> scale-score threshold.
For basis='level' (deferred): the fractional level target (e.g. 4.2).