Glossary: Schema fields, M–R
Candidate terms extracted from every schema field with a doc comment: <Declaration>.<field> as the label, the comment’s first paragraph as the definition. Each is the asset’s own text, verbatim and not curated, and carries the badge “candidate, extracted”. A person promotes one by authoring it. Authored terms are on their owning instance’s page, and those pages, the counts and the sources are linked from the glossary index.
From: bootstrap-tools 44 (SKOS) · cat-harness 2141 (SKOS) · fhir-harness 78 (SKOS) · folio-assistant-core 179 (SKOS) · folio-assistant-sci 13 (SKOS) · smart-base 13 (SKOS).
One of 4 pages of this type, split by the first letter of the label: all parts.
Size: this page holds 660 terms and is 365 KB before compression, fetched in one request, within its budget of 1.0 MB. There is no search index: the filter below runs over this page, and the A–Z bar jumps within it.
| target | mapped | unmapped | undetermined | why undetermined |
|---|---|---|---|---|
fhir | 0 | 2468 | 0 | — |
skos | 0 | 2468 | 0 | — |
Undetermined is never “no match”. A vocabulary that could not be reached has said nothing, and the column above keeps that apart from a checked miss. The counts are reported and not graded: an unmapped candidate may be a term this corpus is right to coin.
Per term. Every term on this page is unmapped on fhir, and unmapped on skos, unless its entry says otherwise. No entry says otherwise.
660 shown
M
- Manuscript.commit_sha candidate, extracted
Git commit SHA from which objects were extracted.
- Manuscript.generated_at candidate, extracted
ISO-8601 timestamp of extraction.
- Manuscript.lean_docs_url candidate, extracted
Base URL for doc-gen4 Lean documentation pages.
- Manuscript.pdf_url candidate, extracted
URL to the published PDF.
- Manuscript.repo candidate, extracted
GitHub repository (owner/name).
- MappingScopeSchema.consulted candidate, extracted
The schemes or collections consulted. **May be empty**, and an empty list is a determined finding — this instance declares nothing to check against for that target — rather than a missing field.
- MappingScopeSchema.unreachable_reason candidate, extracted
Present when the target could not be consulted at all.
- MappingScopeSchema.via candidate, extracted
How they were reached:
localfor in-repo schemes, a host for a service.- MaskedRegionSchema.reason candidate, extracted
Why this region is masked, in a person's words. Never generated.
- MaskedRegionSchema.x1 candidate, extracted
Exclusive.
- MaskedRegionSchema.y1 candidate, extracted
Exclusive.
- MaterialisedDirectory.absPath candidate, extracted
Path the directory was (or would be) created at — inside the instance, or at the repository root for a
repository-scoped entry.- MaterialisedDirectory.created candidate, extracted
True when this call created the directory; false when it already existed.
- MaterialisedDirectory.declaredBy candidate, extracted
The instance whose declaration contributed the entry — may be a dependency.
- MaterialisedDirectory.markerWritten candidate, extracted
True when this call wrote the keep-marker; false when one was already there.
- MaterialisedDirectory.path candidate, extracted
The DECLARED path, as written. Reported so a caller need not recover it from
absPath—relative(instanceRoot, absPath)gives../beansfor a repository-scoped entry, which is a traversal rather than a name.- MaterialisedDirectory.scope candidate, extracted
Which root
pathwas resolved against; absent means the instance's.- MaterializationSchema.bytes candidate, extracted
Bytes held locally. Absent means not measured, which is not zero.
- MaterializationSchema.collectionBytes candidate, extracted
Bytes the WHOLE remote collection holds, where it is known. This is what makes a
sizegate answerable: 3 items of 361.55 GB is a fraction, and "3 items" alone is not.- MaterializationSchema.expiresAt candidate, extracted
When this copy expires. Absent on a
workingcopy is aretentionfinding, not a default of "forever". Absent on anarchivalone is the SPECIFICATION — see {@link freshness}, which reports the two differently.- MaterializationSchema.fixity candidate, extracted
Required when
purposeisarchivalorboth.- MaterializationSchema.inputs candidate, extracted
Required when
purposeiscompiled, and only allowed then. See {@link CompiledInputsSchema}.- MaterializationSchema.lastReadAt candidate, extracted
When this copy was last read. Absent means 'not recorded'.
- MaterializationSchema.localPath candidate, extracted
Where the bytes landed, instance-relative. Present iff
materialized.- MaterializationSchema.materializedAt candidate, extracted
When the local copy was taken, and against what upstream version.
- MaterializationSchema.note candidate, extracted
Why the record cannot say more than it does.
- MaterializationSchema.provenance candidate, extracted
WHERE THIS CAME FROM — see {@link SourceProvenanceSchema}.
- MaterializationSchema.purpose candidate, extracted
Why the bytes were taken. Required on anything materialized — the gates mean different things under each.
- MeasurementSchema.command candidate, extracted
The exact command, runnable as written.
- MeasurementSchema.date candidate, extracted
ISO 8601 date the measurement was taken.
- MeasurementSchema.result candidate, extracted
What it produced — the number, or the summary line.
Changed paths a declared merge pattern would REFUSE — the authored ones (T3, R7).
- MemberFactsSchema.conflictRisk candidate, extracted
Shares an AUTHORED path with another live queued PR — input (d), T3.
- MemberFactsSchema.headShaMatchesCi candidate, extracted
Did CI run on the head that would be merged (T2; bean
u7beitem 3).- MemberFactsSchema.mvp candidate, extracted
Tagged MVP / feature-priority, by epic or label — input (c).
- MemberFactsSchema.overlapKind candidate, extracted
WHICH KIND of collision this PR has with the rest of the live queue.
- MemberFactsSchema.ownCi candidate, extracted
The PR's own CI on its head: T2's first evidence.
- MemberFactsSchema.readiness candidate, extracted
Has the OWNING session signalled this PR is finished, against
main? Beanuoob(merge gate (f)), read byRule_NotReady.readyis the only admitting value, and the others stay apart because each has a different remedy:- MemberFactsSchema.refused candidate, extracted
A declared merge pattern refuses a conflicted path, or GitHub reports it dirty.
- MemberFactsSchema.seedsStaging candidate, extracted
Input (a):
touchesHarnessAND a linked bean descends from the separation goal — the PR seeds the staging repos, which is the reason the owner gave for (a). Seemerge-priority.dmn's note.- MemberFactsSchema.sizeBand candidate, extracted
Changed files plus changed lines, banded — input (d).
- MemberFactsSchema.touchesHarness candidate, extracted
An authored path under
cat-harness/or its tools layer's directory. The raw measurement behind input (a) — NOT input (a) itself: measured 2026-10-02, it was true for 11 of the 11 PRs in that day's queue.Touches a shared declaration: a schema, generator,
<instance>.json, BPMN or DMN (T3, R6).- MemberFactsSchema.unblocks candidate, extracted
Open PRs or beans this unblocks, or tangle edges it removes — input (b).
- MemoryNodeSchema.$schema candidate, extracted
What this file IS, declared inside it.
- MemoryNodeSchema.archived candidate, extracted
Retained as a graph node, injected into no agent's file.
- MemoryNodeSchema.detail candidate, extracted
The part of this entry that is NOT injected — written beside
MEMORY.mdand pointed at, for the agent to read on demand.- MemoryNodeSchema.label candidate, extracted
STABLE, TRAP or BASELINE.
- MemoryNodeSchema.measured candidate, extracted
Provenance for a measured number. **Required when
labelisbaseline**, enforced by the refinement below rather than by the type, so the error names the rule instead of a missing field.- MergeReviewSchema.repository candidate, extracted
The repository
pris in,owner/repo. Optional, aspris: the gate looks a verdict up by its head sha, but a reader of the store cannot link a pull request number without it. A reader links only anowner/repospelling, so a malformed one costs the link, never the verdict.- MethodologyFrontMatterSchema.applies-when candidate, extracted
When this methodology applies, phrased so
methodology-adoption's selection question can reach it — and saying what it is NOT for.- MethodologyFrontMatterSchema.evidence candidate, extracted
The ingested sources, each as
library/<bib-slug>.- MethodologyFrontMatterSchema.name candidate, extracted
Stable id, referenced from skills and diagrams. Kebab-case, like a skill's.
- MethodologyFrontMatterSchema.origin candidate, extracted
Authors, a publication, a standards body.
- MethodologyFrontMatterSchema.title candidate, extracted
What a reader sees. Not the id with capitals — it may name the acronym's expansion.
- MethodologyRecommendationSchema.avoidWhen candidate, extracted
Under what conditions this method should NOT be used. Matched against context.
- MethodologyRecommendationSchema.complexity candidate, extracted
Low / Medium / High.
- MethodologyRecommendationSchema.dataRequirements candidate, extracted
What data the method requires.
- MethodologyRecommendationSchema.family candidate, extracted
Which method family: mcdm-aggregation, probabilistic, social, classical, hybrid-mcdm.
- MethodologyRecommendationSchema.matchTier candidate, extracted
How well this method fits the context. A tier, not a score — the platform refuses pseudo-measurement (methodology-adoption: "never quantify a judgement to make it look measured; where a method scores and sums, take structure and refuse arithmetic").
- MethodologyRecommendationSchema.methodologyId candidate, extracted
The methodology's
namefrom its front matter (kebab-case id).- MethodologyRecommendationSchema.rationale candidate, extracted
Why this methodology fits the given context. Matched against
applies-when.- MethodologyRecommendationSchema.source candidate, extracted
Bibliographic source: paper title, authors, arxiv id or standard.
- MethodologyRecommendationSchema.title candidate, extracted
The methodology's human-readable title.
- MethodologySelectionResultSchema.context candidate, extracted
The context that was provided.
- MethodologySelectionResultSchema.noMatchReason candidate, extracted
If no methodology fits, this explains why. Per methodology-adoption: "If none fits, that is a finding, not a licence."
- MethodologySelectionResultSchema.recommendations candidate, extracted
Ranked recommendations, best fit first.
- MilestonePlanSchema.coveredOpen candidate, extracted
SUMMED from the closures, never
openTotal - orphanOpen.- MilestonePlanSchema.duplicateIds candidate, extracted
Bean ids held by more than one file. A store defect, reported.
- MilestonePlanSchema.orphanOpen candidate, extracted
Open beans with no milestone anywhere above them.
- MilestoneRollupSchema.epics candidate, extracted
Direct children typed
epic.- MilestoneRollupSchema.file candidate, extracted
Repository-relative path of the milestone's bean file.
closed / (closed + open);nullwhen nothing is classified, which is NOT the same as 0 — see the module note onmilestone-rollup.ts.- MilestoneRollupSchema.status candidate, extracted
The milestone's OWN status — not derived from its subtree.
- MilestoneRollupSchema.total candidate, extracted
Transitive descendants, excluding the milestone itself.
- MilestoneRollupSchema.unclassified candidate, extracted
In neither status set — reported alone, folded into neither figure.
- ModelCapabilitiesSchema.capabilities candidate, extracted
Specific named capabilities (e.g. "code_execution", "json_schema").
- ModelCapabilitiesSchema.context_window candidate, extracted
Minimum context window size in tokens.
- ModelCapabilitiesSchema.multimodal candidate, extracted
Whether multimodal input (images, diagrams) is required.
- ModelCapabilitiesSchema.reasoning_tier candidate, extracted
Minimum reasoning tier required.
- ModelCapabilitiesSchema.tool_use candidate, extracted
Whether function calling / tool invocation is required.
- ModelEntrySchema.id candidate, extracted
The model identifier, exactly as the runtime reports it.
- ModelEntrySchema.note candidate, extracted
Anything a reader needs that the fields above cannot carry.
- ModelEntrySchema.preferredLanguages candidate, extracted
BCP 47 tags, best first.
- ModelEntrySchema.title candidate, extracted
Human-readable name, for a report a person reads.
- ModelEntrySchema.validatedBy candidate, extracted
Who checked, when
validationishuman-validated.- ModelEntrySchema.validatedOn candidate, extracted
ISO date of that check. A validation with no date cannot go stale.
- ModelEntrySchema.validation candidate, extracted
How the list above came to be believed. Required — see the header.
- MountConsentSchema.by candidate, extracted
Who consented: a person, by the identity they act under here.
- MountConsentSchema.evidence candidate, extracted
Where the consent was given (an issue comment, a chat session), so it can be read.
- MountConsentSchema.ref candidate, extracted
The pin consented to. Must equal the mount's
ref; a moved pin asks again.- MountConsentSchema.standing candidate, extracted
Present when this pin's consent was written under a STANDING consent (
trust.standing), by the repin tool rather than by the person at this pin. It holds only while that rule is still declared, unchanged.- MountDefaultsSchema.directories candidate, extracted
Directory ids, in this instance's own declaration.
- MountDefaultsSchema.whole candidate, extracted
Mount the WHOLE instance root — every tracked file at the pin, root files included — instead of its declared directories. For an instance a downstream reads as a checkout rather than as graphs:
bootstrap(whosens.jsonldand README sit at its root) andbootstrap-tools(whosepackage.jsonandtsconfig.jsondo), the two git submodules a remote mount replaces (beannn8e, #2462). Locked as one directory, id*.- MountLockSchema.mounts candidate, extracted
The
remoteMountsit was written for, so a changed pin reads as stale.- MountLockSchema.unmounted candidate, extracted
What the closure reached and did NOT mount, with why — so the check, which reads the lock and never the network, cannot report clean over an instance the mount never laid down.
localandskippedare recorded too: they are answers, and a reader should see them as such.- MountOverrideSchema.trust candidate, extracted
Consent for THIS instance at its own pin. Required for an instance reached through a GITLINK: that is another repository at another commit, and the parent mount's consent does not cover it (roast
1ygpL4.2). Ignored for the declared harness and same-tree instances, which the mount's owntrustcovers.- MountOverrideSchema.whole candidate, extracted
Overrides the harness's
mountDefaults.wholeeither way;directoriesis then ignored.- MountSignatureSchema.network candidate, extracted
The trust network the key is listed in, e.g.
gdhcn.- MountSignatureSchema.signedDigest candidate, extracted
What was signed: the pinned declaration's sha256, as the lock records it.
- MountTrustSchema.standing candidate, extracted
A standing consent for this instance: see "Standing consent" above.
N
- NameRuleKind.file candidate, extracted
Absolute path of the module holding the export.
- NamespaceIndexSchema.shards candidate, extracted
Every shard file in this namespace, relative to it. A prefix not listed has no entries, so no request is needed to say so.
- NarrativeMapping.context candidate, extracted
Context in which this mapping is valid.
- NarrativeMapping.latex_label candidate, extracted
LaTeX label (e.g., "def:brings-surface").
- NarrativeMapping.lean_id candidate, extracted
Lean identifier (e.g., "QOU.BringsSurface.BringsSurface").
- NarrativeMapping.narrative candidate, extracted
Narrative string (e.g., "Bring's Surface").
- NarrativeSchema.confirmed_by candidate, extracted
Who ACCEPTED the words. Must be a human; see the module docstring.
- NarrativeSchema.drafted_by candidate, extracted
Who wrote the words. An
agentmust name itsmodel— seeattribution.ts.- NarrativeSchema.lang candidate, extracted
The language of
text, as a BCP 47 tag. Required on every narrative that has text; the emptynot-authoredslot claims no language (owner ruling 2026-10-22, 1A).- NarrativeSchema.rejection_reason candidate, extracted
Why it was turned down. Required when rejected, and never empty.
- NarrativeSchema.text candidate, extracted
The narrative itself, or null when nobody has written one.
- NestedDirectory.id candidate, extracted
<parent id>/<own id>— unique across the instance, and says who declares it.- NestedDirectory.ownId candidate, extracted
The entry's OWN id, as
resolveDirectoriesand a mount marker spell it.- NestedDirectory.subgraph candidate, extracted
truewhen the entry declares itself an instance subgraph (beancmsl).- NodeDependency.package candidate, extracted
npm package name.
- NodeDependency.version candidate, extracted
semver range (e.g., "^1.0.0", ">=2.0.0").
- NodeKind.id candidate, extracted
The kind's NAME — its identity and its URL segment (
changeset). Never versioned.- NodeKind.order candidate, extracted
Every ancestor and then this kind, deepest first: the composition order.
- NodeKind.overrides candidate, extracted
Inherited fields this layer redefines, on purpose.
- NodeKind.own candidate, extracted
The fields this layer itself declares.
- NodeKind.schema candidate, extracted
Every field, composed in {@link order}.
- NodeKind.tag candidate, extracted
<id>/<version>: what a writer stamps in$schema. Absent when unversioned.- NodeKind.version candidate, extracted
SemVer, for a kind whose nodes are files (issue #2195). Absent for a mixin (
themed), which has no$schemaof its own.- NodeKindEntry.declaredBy candidate, extracted
The instance that declares it: the one whose code holds the
nodeKind(). Absent for a kind reached only as an ANCESTOR (a mixin such asthemed): no typology names it, so nothing here says where it lives, and guessing from a child would file it under whichever child was read first.- NodeKindEntry.foundBy candidate, extracted
How its nodes are found when no typology holds them: by a NAME RULE ({@link NAME_RULE_KINDS}).
instance-declarationis each instance's own<instance>.json, which sits at an instance root rather than in a declared directory, so no typology can name it. Absent: found throughholdings.- NodeKindEntry.holdings candidate, extracted
The graph typologies that hold nodes of this kind, and the families they file it under.
- NodeKindEntry.id candidate, extracted
nodeKind()'s id — a$schematag for a tagged kind, a name for a mixin.- NodeKindEntry.module candidate, extracted
Where, repo-relative, and under which export. Absent for an ancestor reached only as a parent.
- NodeKindEntry.pages candidate, extracted
The kind's own page renderer, when its validator node names one (
pages), repo-relative. Absent: the generic pages built from the schema.- NodeKindEntry.parents candidate, extracted
Direct parents, by id, in declared order.
- NodeKindEntry.subclasses candidate, extracted
Direct subclasses, by id, sorted: the kinds that name this one as a parent.
- NodeKindEntry.tag candidate, extracted
<id>/<version>: the$schemaa current node carries.- NodeKindEntry.version candidate, extracted
SemVer, for a kind whose nodes are files; absent for a mixin.
- NodeKindIndex.collisions candidate, extracted
Two different kinds sharing one id — always a finding.
- NotebookSectionSchema.cell_end candidate, extracted
Zero-based index of its last cell, inclusive.
- NotebookSectionSchema.cell_start candidate, extracted
Zero-based index of the section's first cell, which is its heading cell.
- NotebookSectionSchema.code_cells candidate, extracted
How many of those cells are code. Kept verbatim in the section text, never executed.
- NotebookSourceSchema.language candidate, extracted
The kernel language the notebook declares, or null when it declares none.
- NotebookSourceSchema.mimetype_sniffed candidate, extracted
The same two fields every rung writes (bean
nso8), so the L1 gate'stechnical-metadatarequirement reads them as it does for a PDF. A notebook has no magic bytes, so its type is decided by CONTENT, andmimetype_sourcesays exactly that rather than claiming a sniff.- NotebookSourceSchema.nbformat candidate, extracted
A notebook is JSON; this is what the file declared (
nbformat), not sniffed bytes.- NotebookStructureSchema.structure_note candidate, extracted
What the rung did NOT claim. Required when
toc_sourceisnone.- NotebookStructureSchema.toc_source candidate, extracted
headings: sections come from markdown headings.none: the notebook has no markdown heading, so it is one section, andstructure_notesays so.Issues, pull requests, commits. See {@link ArtefactRefSchema}.
Who it is for, or who raised it.
BPMN process ids this belongs to.
Anything else in the graph this points at. See {@link KgRefSchema}.
Role ids — the swimlane this sits in.
scenarios/roles.json.Specific activities, each with its process.
- NpmSourceSchema.integrity candidate, extracted
Integrity SRI string (sha256 or sha512) for fixity verification.
- NpmSourceSchema.package candidate, extracted
Scoped or unscoped package name (e.g.
@litlfred/cat-harness).- NpmSourceSchema.tool candidate, extracted
Declared Tool that retrieves/mounts the package. Default: 'npm'.
- NpmSourceSchema.upstreamPath candidate, extracted
Upstream path inside the extracted package tarball.
- NpmSourceSchema.version candidate, extracted
Exact semver version triple (no ranges).
O
- Occurrence.from candidate, extracted
The instance the FILE belongs to.
- Occurrence.text candidate, extracted
The whole line, which is what an exemption matches against.
- Occurrence.to candidate, extracted
The instance whose NAME was found.
- OdrlPolicySchema.assigner candidate, extracted
An Agreement names who grants it. A Set does not have to.
- OdrlPolicySchema.inheritFrom candidate, extracted
Policies whose rules this one inherits: ODRL's
inheritFrom.- OdrlPolicySchema.profile candidate, extracted
The folio ODRL profile; required, because the actions only mean something against it.
- OdrlRuleSchema.assignee candidate, extracted
An actor id, or {@link ANYONE}.
- OdrlRuleSchema.target candidate, extracted
What the rule applies to; absent means the instance's whole graph.
- OpenApiConfigDocumentSchema.description candidate, extracted
One sentence on what this API is and why this instance holds it.
- OpenApiConfigDocumentSchema.id candidate, extracted
The document's id: its file stem, its page directory and its IRI segment.
- OpenApiConfigDocumentSchema.title candidate, extracted
What a reader calls it. Absent, the document's own
info.titleis used.- OpenApiConfigSchema.directory candidate, extracted
The
<instance>.jsondirectory id of graph typologyopenapithe documents are ingested into.- OpenApiConfigSchema.documents candidate, extracted
Documents read from a repository (
openapi-ingest-source).- OpenApiConfigSchema.rendered candidate, extracted
Documents read from a published spec URL (
openapi-ingest-rendered).- OpenApiOperation.declaredId candidate, extracted
Whether {@link id} is the document's own
operationId.- OpenApiOperation.id candidate, extracted
The node id:
operationId, or<method>-<path>made filename-safe.- OpenApiProvenanceSchema.file candidate, extracted
The document file beside this one.
- OpenApiProvenanceSchema.materialization candidate, extracted
The document IS a held copy of upstream bytes, so it says so in the shared
folio-materialization/v1shape (core'sMaterializationSchema): where from, where it landed, why it was taken, its fixity and the five gates. Checked here only as far as this harness relies on it — the state, the path and the digest; the full record is that schema's to judge, andcheck:read-only-graphsreadsstateto agree with the directory'sreadOnly.- OpenApiProvenanceSchema.openapi candidate, extracted
The document's
openapiversion, and itsinfo.title/info.version, as published.- OpenApiProvenanceSchema.operations candidate, extracted
How many operations it holds — every one of which gets a page.
- OpenApiRenderedSourceSchema.url candidate, extracted
The published spec's URL: an OpenAPI 3 JSON document, served over https.
- OpenApiSourceSchema.path candidate, extracted
Path of the document inside that repository.
- OpenApiSourceSchema.ref candidate, extracted
Branch, tag or commit to read it from. The ingest records the commit it resolved to.
- OpenApiSourceSchema.repository candidate, extracted
owner/repoon GitHub.- OpenInstanceSchema.atNode candidate, extracted
The node the session believes it is at.
- OpenInstanceSchema.instance candidate, extracted
The instance id — the stem of a node in the bean graph's workflow-state.
- OpenInstanceSchema.process candidate, extracted
The BPMN process it walks. Named, so a reader need not open the instance.
- OrderedStep.fatal candidate, extracted
Does a failure here end the run?
- OrderedStep.label candidate, extracted
Free-text, carried through to the report so a reader is not left guessing.
- OrderedStep.needs candidate, extracted
Ids this step must run after. Absent or empty means it may run first.
- OrphanSidecar.kind candidate, extracted
subject.kindandsubject.id, when the sidecar records them.- OrphanSidecar.sidecar candidate, extracted
Repo-relative path of the sidecar itself.
- OrphanSidecar.subject candidate, extracted
subject.pathas the sidecar records it, orundefinedif it records none.- OrphanSidecar.subjectExists candidate, extracted
Whether that subject is on disk.
undefinedmeans the sidecar could not be read or names no path — a THIRD state, kept because "could not tell" rendered as either answer is how the eight fragment sidecars below got the wrong diagnosis in the first place.- Overrule.finding candidate, extracted
{@link Finding.id}.
- Overrule.note candidate, extracted
{@link AuditNote.id}.
- Owned.path candidate, extracted
Path inside the instance, without extension,
/-separated.- Owned.root candidate, extracted
Instance root, absolute.
P
- PackageManifest.apt candidate, extracted
APT packages (Ubuntu/Debian).
- PackageManifest.architectures candidate, extracted
Target CPU architecture(s).
- PackageManifest.baseImage candidate, extracted
Base Docker image. Ubuntu LTS only.
- PackageManifest.description candidate, extracted
What this package provides.
- PackageManifest.envVars candidate, extracted
Environment variables required at runtime.
- PackageManifest.exposedPorts candidate, extracted
Ports exposed by services in this package.
- PackageManifest.healthcheck candidate, extracted
OCI-compatible healthcheck.
- PackageManifest.installers candidate, extracted
Tools installed via custom scripts/curl.
- PackageManifest.name candidate, extracted
Human-readable name (e.g., "authoring-math", "authoring-who-smart-guidelines").
- PackageManifest.node candidate, extracted
Node/Bun packages.
- PackageManifest.ociAnnotations candidate, extracted
OCI image annotations.
- PackageManifest.packageId candidate, extracted
Package identifier matching skills-config.json name or "local".
- PackageManifest.python candidate, extracted
Python packages (installed via uv/pip).
- PagesServingObservationSchema.observedAt candidate, extracted
ISO 8601, when the CI job fetched the site.
- PagesServingObservationSchema.run candidate, extracted
The workflow run that made the observation, for a reader to open.
- PagesServingObservationSchema.site candidate, extracted
The published site's root URL.
- PagesServingUrlSchema.bytes candidate, extracted
Bytes in the response body.
- PagesServingUrlSchema.expectedPage candidate, extracted
Whether the body is the page asked for, rather than Pages' own 404.
- PagesServingUrlSchema.status candidate, extracted
HTTP status of the final response.
- PairAttestationSchema.by candidate, extracted
baselinecounts as a judgement (owner, 2026-10-01): it records the state the drift is measured from.- PanelControl.because candidate, extracted
Why this control exists, for whoever is deciding whether to keep it.
- PanelControl.label candidate, extracted
The accessible name. Words, because a glyph alone is a guess.
- PanelControl.needs candidate, extracted
What the environment must be able to do for this to work.
- Paper.abstract candidate, extracted
Paper abstract (plain text or markdown).
- Paper.chapters candidate, extracted
Chapter references — URIs to chapter .ts manifests.
- Paper.emails candidate, extracted
Author email addresses (parallel to authors array).
- Paper.macros candidate, extracted
Custom LaTeX macros for this paper.
- Paper.urls candidate, extracted
Author URLs / homepages (parallel to authors array).
- PaperMacro.tex candidate, extracted
LaTeX expansion (e.g. "\\mathfrak{p}"). Used by KaTeX and LaTeX.
- PaperMacro.unicode candidate, extracted
UTF-8 display string for plain-text contexts (e.g. "𝔭").
- PaperRef.description candidate, extracted
Optional description shown on the landing page.
- PaperRef.dir candidate, extracted
Directory name under folio/ (e.g. "quantum-observable-universe").
Tags for filtering on the landing page.
- PaperRef.title candidate, extracted
Optional display title override (otherwise read from paper manifest).
- PaperRef.url candidate, extracted
Folio viewer URL for this paper. Used for: - Cross-paper dependency resolution (qualified
usesrefs) - BibTeXfolio-urlfield for folio-hosted citations - Read-only import of external papers for commenting/referencing- PDCrossing.index candidate, extracted
Crossing index.
- PDCrossing.strands candidate, extracted
Four strand indices meeting at the crossing [i, j, k, l].
- PdfDiagnosticsSchema.figure_sequence_gaps candidate, extracted
Numbers missing from a caption run, e.g. "table 2.1" beside a Table 2.2.
- PdfDiagnosticsSchema.latex_overlay candidate, extracted
What
latex-math-overlay.pydid: the main.texfile it read, how many sections it found there and how many PDF sections took their text from it, the three gates it applied, and a per-section record of each match.- PdfDiagnosticsSchema.page_label_conflicts candidate, extracted
Pages where label sources disagree, e.g. /PageLabels "3" against a printed "iii".
- PdfDiagnosticsSchema.toc_alignment candidate, extracted
Where a printed contents page and the body disagree (issue #2302) — drafts drift. Present only when the document has a contents page. Each list is capped;
countis the full number.Share of the sections an inferred TOC would cut that hold under 50 characters, front matter excluded — the third trust test (issue #2302): above 25% (from 5 sections up) the tree is refused as headings without bodies. Reported whether or not it passed.
- PdfDiagnosticsSchema.toc_inferred_mean_confidence candidate, extracted
Mean confidence of the inferred entries, the input to the second trust test (issue #2302): below 0.6 the inference is not used and the entry is split one section per page. Reported whether or not it passed.
- PdfDiagnosticsSchema.toc_inferred_method candidate, extracted
Which inference produced an inferred TOC (issue #2302): a printed contents page, heading styles read from font metrics, the text-pattern heuristic that OCR'd text falls back to, or
latex: the\sectioncommands of the paper's own LaTeX source, each placed on its PDF page (cat-harness-tools_latex_headings.py).- PdfMetadataSchema.title_correction candidate, extracted
An EDITORIAL correction, outranking every extracted title (bean
w6fu, step b). Authored, never produced by a rung, and carried over by a re-ingest.basissays what the title was read from, so it can be checked rather than trusted.- PdfMetadataSchema.title_evidence candidate, extracted
Every candidate the resolver saw, and which sources agreed — the record an editor decides from.
- PdfMetadataSchema.title_raw candidate, extracted
The text walk's title before resolution, kept so a re-run resolves from it rather than from its own answer.
- PdfMetadataSchema.title_source candidate, extracted
Where
titlecame from (beanw6fu,scripts/_pdf_title.py): the PDF'sTitlemetadata, its first-page heading or its outline when another source CORROBORATES it, else the text walk's title (text) or nothing (filename, and the manifest falls back to the entry id).- PdfMetadataSchema.title_verified candidate, extracted
False when no source corroborated the title: the raw title was kept, not guessed at.
- PdfSectionSchema.label_start candidate, extracted
Printed labels of page_start / page_end, where known (issue #2302).
- PdfSourceSchema.math_source candidate, extracted
Where the entry's MATHEMATICS comes from, written by
latex-math-overlay.py:latexwhen at least one section's text was replaced by the paper's own LaTeX,text-layerwhen the overlay ran and none passed its gates. Absent when the overlay never ran.- PdfSourceSchema.runContext candidate, extracted
WHICH SOFTWARE produced the text, cited by digest: the
folio-run-context/v1of the run that wrote it (schemas/run-context.ts), reused rather than restated — the one optionalrunContext: { sha256 }reference thatdocs/proposals/tool-releases-2026-10-07.mdgives every run and report record (owner, C1 = a).- PdfStructureSchema.container candidate, extracted
A bound volume split into its papers by qou's
split-bound-volume.py: the volume keepscontainer: true, names the per-paper entries incontainer_childrenand says why incontainer_note. Ranking and the library index skip a container, so its pages are not counted twice.- PdfStructureSchema.duplicate_of candidate, extracted
The library holds some works more than once (a preprint beside its published version, one PDF under two names). The duplicate copy names the entry to prefer in
duplicate_ofand says why induplicate_note; the preferred entry lists its copies induplicate_copies. Adjudicated in qoudocs/audits/library-duplicate-ledger.json.- PdfStructureSchema.figures candidate, extracted
The list of figures and tables, cross-checked (issue #2302).
- PdfStructureSchema.granularity candidate, extracted
page: written bypdf-pages.py, one section per page.slide: written byslides-structure.py, one section per slide (beanscfh).- PdfStructureSchema.pages candidate, extracted
Every physical page with its printed label (issue #2302).
- PdfStructureSchema.structure_note candidate, extracted
What a rung did NOT claim. Required by check-l1-complete when there are no sections.
- PdfStructureSchema.text_extraction candidate, extracted
How the text was obtained when it was done by hand, outside the producer: the route, why, which engine and settings, and how far the result can be trusted. Written for scans OCR'd before
source.runContextexisted; a producer run recordssource.runContextinstead.- PdfStructureSchema.toc_note candidate, extracted
Where the headings came from when a person supplied them, not the producer.
- PdfStructureSchema.toc_undetermined_reason candidate, extracted
Why an inferred TOC was not trusted, in a sentence a person can check.
- PdfTocEntrySchema.confidence candidate, extracted
Inferred entries only (issue #2302): how sure the inference is, 0..1, and which independent evidence agreed — a printed contents page, the entry found in the body, a heading style, a section number, another extractor. Null on an outline entry, which is not scored.
- PdfTocEntrySchema.page_label candidate, extracted
The printed label of
page("iv", "23"), where known;pageis physical.- PermissionDef.includedIn candidate, extracted
The broader actions this one is part of: another action here, or one of ODRL's common vocabulary (
schemas/odrl.ts#ODRL_ACTIONS). This is what makes the vocabulary an ODRL profile (issue #1180): a permission to a broader action permits every action included in it. Required and non-empty, so every action has a way up toodrl:use.- PermissionDef.title candidate, extracted
Display text and the sentence under it —
schemas/kg-node.ts, like every node.- PermitRequest.actor candidate, extracted
The actor asking, or
nullfor an unauthenticated reader.- PermitRequest.actorRoles candidate, extracted
The roles the actor may act as (
ActorDef.roles); empty or absent means unconstrained.- PinnedConceptSchema.deprecated candidate, extracted
The edition itself marks the code deprecated: still valid to read, not to write anew. Absent means the edition says nothing, not "current". First user: the SPDX License List (
isDeprecatedLicenseId), beansd5v.- PinnedConceptSchema.display candidate, extracted
The human label, and the ONLY field a glossary candidate can match on.
- PinnedConceptSchema.system candidate, extracted
The code system's id inside the IG — its identity there, not a URL.
- PinnedTerminologySchema.pin candidate, extracted
The external-schema record this version is read from.
- PinnedTerminologySchema.version candidate, extracted
The pinned version, which must equal that record's.
- PipelinePluginContribution.implementation candidate, extracted
What fills it. Shape is fixed per kind by
pipeline/pipeline-plugins.ts.- PipelinePluginContribution.kind candidate, extracted
The slot this fills, e.g.
"lean-lexer". One contributor per kind.- PipelinePluginNodeSchema.implementation candidate, extracted
The table of slot implementations, typed against the slots' contract.
- PipelinePluginNodeSchema.slot candidate, extracted
The generic pipeline slot this fills (
pipeline/pipeline-plugins.ts); the key into the table.- PlacementSchema.coverage candidate, extracted
Placed area over page area, AT THIS PLACEMENT. Sizes may differ per use.
- PlacementSchema.imagesOnPage candidate, extracted
How many images share this page.
- PlacementSchema.note candidate, extracted
What this copy serves HERE — never what the image is, which is the narrative's job. "a recurring header for principle 8" belongs here; "an irregular grid of coloured squares" belongs on the image.
- PlacementSchema.page candidate, extracted
The page it sits on, 1-based as a reader counts.
- PlanarDiagram.crossing_count candidate, extracted
Number of crossings.
- PlanarDiagram.hyperbolic_volume candidate, extracted
Hyperbolic volume (if known).
- PlanarDiagram.name candidate, extracted
Knot name (e.g., "4_1" for figure-eight).
- PlanarDiagram.volume_error_bound candidate, extracted
Error bound on the volume.
- PrComment.association candidate, extracted
GitHub's
author_association: OWNER, MEMBER, COLLABORATOR, CONTRIBUTOR, NONE… Absent when not known.- precisionScalar.dps candidate, extracted
Significant decimal digits in
value.- precisionScalar.value candidate, extracted
Signed decimal: optional
-, integer part, optional.<frac>, optional[eE][+-]?<int>exponent. Includes uppercaseEand explicit positive sign on the exponent — both produced by various MPFR / mpmath formatting paths.- PredictionCheck.confirmed candidate, extracted
Predicted and measured.
- PredictionCheck.missed candidate, extracted
Measured to change but not predicted: a cone defect, never a pass.
- PredictionCheck.unconfirmed candidate, extracted
Predicted, not measured: expected for content pages that load their data (rule 3), suspicious otherwise.
- PrintGeometrySchema.fontScale candidate, extracted
Body scale relative to the format, as a unitless multiplier.
- PrintGeometrySchema.height candidate, extracted
Trim height, as a CSS length —
29.7cmfor A4.- PrintGeometrySchema.margin candidate, extracted
Page margin, as a CSS length.
- PrintGeometrySchema.width candidate, extracted
Trim width, as a CSS length —
21cmfor A4.- ProcessTask.process candidate, extracted
The
<bpmn:process id>.- ProcessTask.task candidate, extracted
The task's id in that process.
- ProofBlock.computation candidate, extracted
Optional computational verification (Python/SnapPea/sympy/etc.).
- ProofBlock.computation candidate, extracted
Optional computational verification (Python/SnapPea/sympy/etc.).
- ProofBlock.lean candidate, extracted
Optional Lean formalization of the proof.
- ProofBlock.lean candidate, extracted
Optional Lean formalization of the proof.
- ProofBlock.of candidate, extracted
Label of the provable block this proof establishes (reverse link of
ProvableBase.proofs[]). Canonical field name — use in preference toproves.- ProofBlock.of candidate, extracted
Label of the provable block this proof establishes (reverse link of
ProvableBase.proofs[]). Canonical field name — use in preference toproves.- ProofGoal.hypotheses candidate, extracted
Local context: hypotheses available.
- ProofGoal.index candidate, extracted
Goal index (0-based).
- ProofGoal.target candidate, extracted
The type to be proved (the goal).
- ProofHypothesis.is_user_introduced candidate, extracted
Whether this hypothesis was introduced by the user.
- ProofHypothesis.name candidate, extracted
Hypothesis name.
- ProofHypothesis.type candidate, extracted
Hypothesis type.
- ProofObject.label candidate, extracted
LaTeX label following the project labeling convention.
- ProofObject.uses candidate, extracted
Labels of objects this one depends on (from \uses{}).
- ProofStateExport.column candidate, extracted
Column number.
- ProofStateExport.current_tactic candidate, extracted
Current tactic (if in tactic mode).
- ProofStateExport.declaration candidate, extracted
Fully qualified declaration name.
- ProofStateExport.exported_at candidate, extracted
Timestamp of export.
- ProofStateExport.file candidate, extracted
File path relative to lean/.
- ProofStateExport.goals candidate, extracted
Outstanding goals.
- ProofStateExport.is_complete candidate, extracted
Whether the proof is complete (no remaining goals).
- ProofStateExport.line candidate, extracted
Line number in the Lean file.
- ProposedAction.bean candidate, extracted
Beans id once the action has been queued as work (e.g.
qou-f4el).- ProposedAction.rationale candidate, extracted
Why this action, specifically — the evidence for the proposal.
- ProposedAction.target candidate, extracted
Block label, Lean declaration, or script path the action targets.
- ProseBlock.label candidate, extracted
Optional, and free-form: prose labels are not required to carry a
kind:prefix, unlike every other labelled block.- ProvableBase.examples candidate, extracted
Labels of example blocks illustrating this result.
- ProvableBase.lean candidate, extracted
Optional Lean formalization of the result.
- ProvableBase.proofs candidate, extracted
Labels of proof blocks for this result (like examples[]).
- ProvActivitySchema.cat-harness:underPolicy candidate, extracted
The ODRL policy uid(s) the run was under. Required: a run under no policy cannot be checked. An array when several policies were in force, which is how JSON-LD writes several values of one property:
decideevaluates every policy an instance holds, so naming only one would misstate what the verdict was computed against.- Publication.host candidate, extracted
WHAT KIND of host serves the renderings. Optional: absent is a third state — the deployment has not said, NOT
github-pages.- Publication.state candidate, extracted
WHAT STATE this instance's publication is in.
draft, always, today.- PublicFieldsSchema.discussion candidate, extracted
GitHub issue holding this comment's discussion, once one is opened.
- PublicFieldsSchema.labels candidate, extracted
Categorisations the intake log carried, keyed by its column header and kept verbatim: a theme, a stakeholder type, a review question, a priority, a committee routing. Read, never re-coded, so a log can grow a column without a schema change.
- PublicFieldsSchema.ref candidate, extracted
"PC-0042": the number a committee and a commenter refer to.
- PublicFieldsSchema.text candidate, extracted
The comment as written.
- PublishedSourceSchema.read_from candidate, extracted
The repo-relative file the identity above was read from.
- PublishedSourceSchema.url candidate, extracted
Where a reader goes to read it — the publisher's own address.
- PythonDependency.package candidate, extracted
PyPI package name.
- PythonDependency.version candidate, extracted
PEP 440 version specifier (e.g., ">=1.0,<2.0").
- PythonDepSchema.distribution candidate, extracted
The distribution name pip installs, e.g.
pdfminer.six.- PythonDepSchema.imports candidate, extracted
The module name Python imports, when it differs from the distribution.
- PythonDepSchema.transitive candidate, extracted
Nothing imports it directly; it is required by something that does.
- PythonDepSchema.why candidate, extracted
Why this is declared — and for a
transitiveentry, why it has no import of its own.
Q
- QaCheckerContribution.check candidate, extracted
The checker itself.
- QaCheckerContribution.criterion candidate, extracted
The criterion id this answers. Must be a criterion the sweep knows.
- QaCheckerContribution.sourceFile candidate, extracted
The file that DEFINES
check, relative to the contributor's own root.- QaCheckerNodeSchema.check candidate, extracted
The dispatch table holding the checker. Its path is the
sourceFilefreshness is hashed over.- QaCheckerNodeSchema.criterion candidate, extracted
The QA criterion this checker answers; the key into the table
checknames.- QaCriterionDefinition.adapters candidate, extracted
Which content adapters this criterion applies to.
- QaCriterionDefinition.also_invalidated_by candidate, extracted
Extra files that invalidate a cached verdict WITHOUT gating applicability — the freshness half of
depends_onon its own.- QaCriterionDefinition.applies_to candidate, extracted
Optional gating: block kinds the criterion applies to. Empty / undefined means "all kinds".
- QaCriterionDefinition.automated candidate, extracted
Whether a deterministic script can run this criterion (true) or it requires agent / human adjudication (false).
- QaCriterionDefinition.checker_contributed candidate, extracted
This criterion is declared here, but its checker belongs to a dependency.
- QaCriterionDefinition.default_severity candidate, extracted
Default severity when the criterion fails.
- QaCriterionDefinition.description candidate, extracted
Human-readable one-liner.
- QaCriterionDefinition.domain candidate, extracted
Domain bucket —
voice,fit,framework,wall, etc.- QaCriterionDefinition.extra_inputs candidate, extracted
Repo-relative paths to extra inputs the checker consults beyond the block under audit — for example, cached audit witnesses (
docs/audits/*.json), the bibliography database (content/schema/references.ts), or cross-block manifests. Concatenated content of these files is hashed intodeps_hashon the entry. Changes invalidate the entry.- QaCriterionDefinition.id candidate, extracted
Stable identifier (e.g.
voice-status-leak).- QaCriterionDefinition.lean_granularity candidate, extracted
How finely this criterion depends on its
.leanfile.- QaCriterionDefinition.profiles candidate, extracted
Which content **profiles** this criterion applies to.
- QaCriterionDefinition.source_file candidate, extracted
Path (repo-relative) to the source file containing this criterion's checker function. Required for
automated: truecriteria; ignored otherwise. Used to compute the entry'sscript_hashso a checker-logic change auto-invalidates downstream sidecar entries.- QaCriterionDefinition.subject candidate, extracted
What the criterion is evaluated **against** — and therefore what shape its checker takes.
- QaCriterionDefinition.untainted candidate, extracted
How an UNTAINTED verification of this criterion is dispatched — and, more to the point, what each dispatched party is NOT allowed to see.
- QaCriterionDefinition.voices candidate, extracted
Named editorial voices this criterion belongs to.
- QaCriterionEntry.evidence candidate, extracted
Evidence for fail/warn verdicts: either a free-form "file:line + verbatim quote" string, or a structured list of { line, text } locations. Some agent reviewers (voice axis especially) emit the structured shape; the live corpus carries both.
- QaCriterionEntry.field_hash candidate, extracted
Source-file hashes captured at audit time.
- QaCriterionEntry.metrics candidate, extracted
Optional structured numeric/heuristic measures a checker emits alongside its verdict — e.g. the detangler axis records per-block graph metrics (
fwd_emitted,out_degree,in_degree,cone_size,edge_span_max,graph_energy,topic_*). Unlikescore(a quality rubric to improve over time) these are descriptive structural measures of the block's position in the dependency graph. Keys are checker-defined; values are scalars or short strings (e.g. a worst-offending target label). Persisted so the sidecar carries the heuristic snapshot, not just pass/fail.- QaCriterionEntry.notes candidate, extracted
Free-form notes.
- QaCriterionEntry.rebuttal candidate, extracted
What rebuts the objection, or "none — survived because …". For scope:"structural" this must name the proved invariant.
- QaCriterionEntry.referee_argument candidate, extracted
The objection in the referee's voice (1-3 sentences). On a da-referee-verdict rollup entry, the strongest objection.
- QaCriterionEntry.result candidate, extracted
Outcome.
- QaCriterionEntry.reviewed_at candidate, extracted
When the audit ran (ISO-8601 UTC).
- QaCriterionEntry.reviewed_sha candidate, extracted
Repo HEAD at audit time (full git SHA). Producers must set this on every new entry; optional in the type only because legacy agent entries (pre-2026-06) omit it — matching the interchange schema (
tools/block-qa-schema/), where it is likewise optional.- QaCriterionEntry.reviewer candidate, extracted
Reviewer identity + provenance.
- QaCriterionEntry.ruling candidate, extracted
Adjudicator ruling on a single objection.
- QaCriterionEntry.scope candidate, extracted
Refutation scope (AGENTS.md "Before declaring refuted").
- QaCriterionEntry.score candidate, extracted
Optional rubric score for rater-style (quality) criteria (
proof-rater-*).value∈ [0, max];rubricholds per-dimension sub-scores. A score is a *quality* measure to improve over time, not a pass/fail gate. Convention: value/max ≥ 0.66 ⇒ result "pass", 0.33–0.66 ⇒ "warn", < 0.33 ⇒ "fail" (severity "minor"). Populated bykind: "agent"reviewers; absent for binary criteria.- QaCriterionEntry.severity candidate, extracted
Severity if
resultis"fail"or"warn".- QaCriterionEntry.verdict candidate, extracted
Block-level rollup verdict. Set ONLY on da-referee-verdict entries.
- QaFieldHash.graph candidate, extracted
Hash of the chapter's
uses[]edge set, for criteria whose verdict is a property of the GRAPH rather than of this block's own files.- QaFieldHash.lean_statement candidate, extracted
Hash of only the **declaration signatures** in the
.leanfile — everything up to each declaration's:=/where.- QaIndexSchema.badges candidate, extracted
Keyed
<nodeId>.<family>orpage.<family>; may be empty.- QaIndexSchema.corpus candidate, extracted
Whether the build that wrote this had the derived QA corpus (bean
4l4d).absent: every badge on the page reads "not available in this build", andbadgesandunsweptare empty because nothing was known. Optional only for an index written before that bean, which the painter reads aspresentwith nounsweptlist — what it then meant.- QaIndexSchema.page candidate, extracted
The page's slug.
- QaIndexSchema.unswept candidate, extracted
Keys whose subject has NO sidecar — "not swept", painted inert. Decided here rather than in the committed page since bean
4l4d.- QaReportEntrySchema.details candidate, extracted
Free-form, because upstream's is. Carried through rather than parsed.
- QaReportFileSchema.status candidate, extracted
Upstream uses
success,error,skipped_existing,updated, … — an OPEN set, so a string. Narrowing it here would reject a real report the day upstream adds a case.- QaReportSchema.phase candidate, extracted
Upstream's own field:
preprocessing,postprocessing, or a build phase.- QaReportSourceSchema.commit candidate, extracted
The commit under test, if known.
- QaReportSourceSchema.repository candidate, extracted
The repository the run happened in, if known.
- QaReportSourceSchema.runUrl candidate, extracted
A URL for the run, if one exists.
- QaReportSourceSchema.tool candidate, extracted
The script or binary, as invoked.
generate_valueset_schemas.py,publisher.jar.- QaResult.families candidate, extracted
Finding families, keyed by name.
- QaResult.producer candidate, extracted
What produced this, so a finding can be traced to the code that made it.
- QaResult.subject candidate, extracted
The artefact the findings are ABOUT — not an authored subject.
- QaResult.total candidate, extracted
Total findings across every family, so "clean" is one read.
- QaResultFamily.count candidate, extracted
How many findings —
0is a determined empty, not an absent answer.- QaResultFamily.entries candidate, extracted
The findings themselves, in whatever shape the producer records.
- QaResultFamily.summary candidate, extracted
What this family means, for a reader who has only the file.
- QaReviewer.actor candidate, extracted
The declared ACTOR this reviewer was acting as — an id in
cat-harness/scenarios/actors/.- QaReviewer.agent_date candidate, extracted
ISO date of the agent review.
- QaReviewer.agent_model candidate, extracted
LLM model identifier, e.g. "claude-opus-4-7", "gemini-2.5-pro".
- QaReviewer.agent_session candidate, extracted
Claude Code session ID or equivalent for the reviewing agent.
- QaReviewer.agent_skill candidate, extracted
Skill that dispatched the review, e.g. "local/one-voice-audit", "local/devils-advocate-watcher".
- QaReviewer.def_hash candidate, extracted
12-char SHA-256 prefix over the criterion DEFINITION's run-affecting fields —
profiles,adapters,applies_to,depends_on,default_severity,lean_granularity. Mismatch ⇒ stale.- QaReviewer.deps_hash candidate, extracted
12-char SHA-256 prefix of the concatenated content of every extra input the checker consults beyond the block under audit (see
QaCriterionDefinition.extra_inputs). Mismatch ⇒ stale.- QaReviewer.id candidate, extracted
Stable identifier — script path, agent name, GitHub login.
- QaReviewer.script_commit_sha candidate, extracted
Full git SHA at which the script was most recently committed at audit time (
git log -n 1 --format=%H -- <source_file>). Used for audit trail and provenance — NOT used to drive staleness (usescript_hashfor that; an uncommitted edit can change the hash without changing the commit SHA).- QaReviewer.script_hash candidate, extracted
12-char SHA-256 prefix of the script's source-file content at the time this entry was written. Populated for
kind: "script". On freshness check, current file hash is compared against this; mismatch ⇒ entry is stale (the checker logic has changed).- QaReviewer.version candidate, extracted
Optional version pin (script revision, model id).
- QaScriptSidecar.$schema candidate, extracted
Schema marker — value:
qa-script/v1.- QaScriptSidecar.criterion_id candidate, extracted
Criterion id this sidecar tracks (e.g.
proof-no-bare-sorries).- QaScriptSidecar.deps_hash candidate, extracted
12-char SHA-256 of concatenated extra_inputs (omitted if none).
- QaScriptSidecar.engine_version candidate, extracted
Engine fingerprint — e.g.
bun-1.3.11+node-22.- QaScriptSidecar.extra_inputs candidate, extracted
Optional repo-relative paths to extra-input files.
- QaScriptSidecar.last_run_at candidate, extracted
ISO-8601 UTC timestamp of the most recent qa-sweep run.
- QaScriptSidecar.last_run_sha candidate, extracted
Repo HEAD at the most recent sweep (full git SHA).
- QaScriptSidecar.script_commit_sha candidate, extracted
Full git SHA of the most recent commit touching
source_file.- QaScriptSidecar.script_hash candidate, extracted
12-char SHA-256 prefix of the source file's content.
- QaScriptSidecar.source_file candidate, extracted
Path (repo-relative) to the checker's source file.
- QaTranslationPagesSchema.pages candidate, extracted
Page slugs (path,
/→-). Empty withcorpus: "absent".- QaWitness.at candidate, extracted
When the review ran (ISO-8601), or absent when the sidecar records none.
- QaWitness.changed candidate, extracted
On
stale, the subject files whose hash moved. Onunknown, why the comparison could not be made, one phrase per unresolved input.- QaWitness.depsHash candidate, extracted
Hash of every extra input the checker consulted beyond the subject.
- QaWitness.freshness candidate, extracted
Whether the verdict still holds for the files on disk.
- QaWitness.id candidate, extracted
Script path, model + skill, or GitHub login.
- QaWitness.kind candidate, extracted
script— a deterministic checker.agent— an LLM.human— a person.- QaWitness.method candidate, extracted
How the verdict was measured, where the checker records it.
- QaWitness.model candidate, extracted
Agent provenance.
- QaWitness.notCompared candidate, extracted
On
partial, the derived inputs that were not re-computed.- QaWitness.scriptHash candidate, extracted
Hash of the checker's own source, and the commit it was last changed in.
- QaWitness.sha candidate, extracted
Repo HEAD at review time.
R
- ReaderFilter.kinds candidate, extracted
Kinds to keep. OR within the list.
- ReaderFilter.properties candidate, extracted
Kind properties to keep, keyed by property name — the owner's *"filter out by kind properties"*. OR within a property's values, AND across properties.
- RecommendationBlock.about candidate, extracted
Labels of the health interventions (DAK adapter) this recommendation is about.
- RecommendationBlock.strength candidate, extracted
Its strength or requirement level, as a code in a declared list.
- RecommendationSchema.by candidate, extracted
GitHub login of the committee member.
- RecommendationSchema.url candidate, extracted
The issue or PR comment it was given in.
- RedFlagCategoryDefinition.definition candidate, extracted
What this category means and what kind of defect it covers.
- RedFlagCategoryDefinition.historicalExample candidate, extracted
Concrete historical example drawn from this repository's history.
- RefactorStrategy.caveat candidate, extracted
Why NOT to apply this in some cases. Populated where known.
- RefactorStrategy.description candidate, extracted
One-line statement of the rewrite.
- RefactorStrategy.example candidate, extracted
Illustrative before/after, always populated — it is the doc.
- RefactorStrategy.expected_gain candidate, extracted
Expected gain, as a rough multiplier on the affected span (0.5 = "about half the length/cost"). Indicative only; the measured
proof-compile-costmetric is the real evidence.- RefactorStrategy.id candidate, extracted
Stable kebab-case id, e.g.
drop-by-exact-wrapper.- RefactorStrategy.lean candidate, extracted
Lean versions this is known-good for. An empty/absent range means UNKNOWN, not "all versions" — the query treats it as non-applicable under strict gating, because an unbounded claim is exactly the assumption that breaks on upgrade.
- RefactorStrategy.match candidate, extracted
Regex (JS syntax, source form) matching the shape to rewrite. Optional: some strategies are structural and only an agent can recognise them.
- RefactorStrategy.replace candidate, extracted
Replacement template, when the rewrite is purely mechanical.
- RefactorStrategy.source candidate, extracted
Provenance — paper, PR, or observation that produced it.
- RefactorStrategy.superseded_by candidate, extracted
For
deprecated: the strategy that replaces this one.- RefactorStrategyFile.$schema candidate, extracted
Schema marker — value:
refactor-strategy/v1.- RefactorStrategyFile.family candidate, extracted
Grouping label, e.g.
mechanical,structural.- ReferenceExemption.file candidate, extracted
Matches the FILE the occurrence sits in, repo-relative. Omitted means any file.
- ReferenceExemption.pattern candidate, extracted
Matches the LINE the occurrence sits on. Omitted means any line, and then
filemust be set.- ReferenceExemption.reason candidate, extracted
Why this is not a reference to the instance. Required.
- ReferenceExemption.to candidate, extracted
Limit to one target instance; omitted means any.
- ReferenceNodeSchema.derivedFrom candidate, extracted
The asset it was taken from: a DC record, a library manifest, a paper's references section. Repo-relative path or IRI.
- ReferenceNodeSchema.heldAt candidate, extracted
The library asset that holds a copy of the work, when one is held. Repo-relative path or IRI.
- ReferenceNodeSchema.provenance candidate, extracted
Where the reference came from. Absent means
authored.- RefSchema.commit candidate, extracted
The commit it resolved to;
nullfor the uncommitted working tree.- RefSchema.ref candidate, extracted
What was asked for —
origin/main, a branch, orworktree.- RefWindowObjectSchema.closedAt candidate, extracted
When the steward actually performed the single write, if it has.
- RefWindowObjectSchema.expires candidate, extracted
The hard close. Set once, at open: there is no extend, because a window that extends on every arrival never closes.
- RefWindowObjectSchema.handoff candidate, extracted
Who takes this window if the holder stops — the field that makes a handover safe. Without it the next steward cannot tell "mine to close" from "someone else's, in flight", and both wrong answers push twice.
- RefWindowObjectSchema.heldBy candidate, extracted
Who holds the window — a session URL,
owner, or an actor id.- ReleaseAssetSchema.bytes candidate, extracted
Size in bytes. Required — this is half of what the node is for.
- ReleaseAssetSchema.contentType candidate, extracted
Media type, where the publisher declares one.
- ReleaseAssetSchema.digest candidate, extracted
sha256:<hex>as GitHub reports it; absent on assets uploaded before GitHub computed digests.- ReleaseAssetSchema.dispositionReason candidate, extracted
Why, for a disposition somebody DECIDED. Required for those; refused for the rest.
- ReleaseAssetSchema.fetchedFrom candidate, extracted
Where the bytes are fetched from. **Required, even for a purged asset** — a record of a file it cannot get back is a size and a hash with no referent.
- ReleaseAssetSchema.name candidate, extracted
The file's published name, e.g.
package.tgz.- ReleaseAssetSchema.url candidate, extracted
Where to fetch the bytes: GitHub's
browser_download_url.- ReleaseIdentitySchema.id candidate, extracted
The release's own id in its origin — a GitHub release id, a registry key.
- ReleaseIdentitySchema.publishedAt candidate, extracted
When it was published.
- ReleaseIdentitySchema.tag candidate, extracted
The git tag, where the origin uses one.
- ReleaseIdentitySchema.url candidate, extracted
The human-facing page for the release, if there is one.
- ReleaseIdentitySchema.version candidate, extracted
The version released. Required: an id alone collides across re-publications.
- ReleaseIris.agent candidate, extracted
<iriBase><version>/— what a program reads.- ReleaseIris.human candidate, extracted
<iriBase>v<major>/— what a person reads.- ReleaseIris.iriBase candidate, extracted
The declared
iriBase, ending in/.- ReleaseIris.major candidate, extracted
MAJORalone, as a number.- ReleaseIris.version candidate, extracted
The declared version,
MAJOR.MINOR.PATCH.- ReleaseOriginRefSchema.repository candidate, extracted
The repository or registry, e.g.
<owner>/<ig-repository>.- RelevanceAssessment.assessed_at candidate, extracted
ISO 8601 timestamp.
- RelevanceAssessment.assessed_by candidate, extracted
Who produced the assessment.
- RelevanceAssessment.human_adjudicated candidate, extracted
Human review of an agent assessment, same semantics as
VerificationEntry.human_adjudicated.- RelevanceAssessment.keyResults candidate, extracted
Results worth consuming. Empty for
tangential/not-relevant.- RelevanceAssessment.proposedActions candidate, extracted
What the author could do about it.
- RelevanceAssessment.rationale candidate, extracted
Why this verdict — the reasoning, not a restatement of the verdict.
- RelevanceAssessment.source_sha candidate, extracted
12-char SHA-256 prefix of the source file at assessment time. A re-upload changes the hash and marks the assessment stale — the same staleness idiom the
.qa.jsonsidecars use.- RemarkBlock.computation candidate, extracted
Optional computational verification (Python/SnapPea/sympy/etc.).
- RemarkBlock.interprets candidate, extracted
Label of the provable block this remark interprets.
- RemarkBlock.lean candidate, extracted
Optional Lean reference (unusual but allowed).
- RemarkSchema.interprets candidate, extracted
Label of the provable block this remark interprets.
- RemedyMatch.invoke candidate, extracted
That Tool's command, so the answer is runnable without a second lookup.
- RemedyMatch.needed_by candidate, extracted
The Tool that needed the host.
- RemedyMatch.none candidate, extracted
Stated: no workaround, and why.
- RemedyMatch.tool candidate, extracted
The Tool to reach for, when there is one.
- RemoteGraph.graphTypologies candidate, extracted
Which parts of the knowledge graph live there.
- RemoteGraph.id candidate, extracted
Stable identifier, unique within an instance — as on a directory, and for the same reason.
- RemoteGraph.url candidate, extracted
Where it is. A reader may follow this; a consumer wanting its bytes may not, without the gates.
- RemoteMountSchema.harness candidate, extracted
The harness's instance name, as its own declaration states it.
- RemoteMountSchema.overrides candidate, extracted
Per-instance overrides across the closure, keyed by instance name.
- RemoteMountSchema.trust candidate, extracted
What makes this mount trusted: a person's consent for THIS pin, or a signature in a declared trust network (
schemas/mount-trust.ts, beanieum, rule H8). Absent means unsigned and unconsented, and a non-staging mount is then refused rather than fetched.- RemotePackageRefSchema.sync candidate, extracted
OPTIONAL, because a wrapper whose only job is to supply Docker requirements should not have to claim a sync strategy to be valid. Required until 2026-09-19, which is why both wrappers carry one.
- RenderedAsset.blockIndex candidate, extracted
0-based index of the ```tex block this replaces in the .md file.
- RenderedAsset.hash candidate, extracted
Content hash for cache invalidation.
- RenderedAsset.mime candidate, extracted
MIME type of the rendered output.
- RenderedAsset.url candidate, extracted
URL or relative path to the rendered file.
- RenderedFacts.files candidate, extracted
The predicted files and undetermined inputs, every renderer's, each pinned.
- RenderedFacts.measured candidate, extracted
The build's measurement, when it was measured.
- RenderedFileSchema.anchors candidate, extracted
The fragment ids within
paththe change alters, for a page that assembles many units (a document page with an anchor per block). A review list linkspath#anchorfor each. Absent when the whole file is the unit, or when a build diff cannot say which part changed.- RenderedFileSchema.hash candidate, extracted
What a review of this file is pinned to (see "A PIN"): for a
conefile, a hash of the blobs of the changed inputs onvia; for abuild-difffile, its content hash. Absent when nothing pinned it.- RenderedFileSchema.path candidate, extracted
Path in the BUILT site, relative to its root, e.g.
ast/artifact/PlanDefinition-X.html.- RenderedFileSchema.via candidate, extracted
How the change reached this file: the changed input first, then each node the cone went through, e.g.
["input/fsh/x.fsh", "IMMZD18SBCG", "PlanDefinition/IMMZD18SBCG"]. Empty for a build diff, which measures that a file changed and cannot say why.- RenderedImpactSchema.inputs candidate, extracted
The changed input files the renderer was given, relative to its source root.
- RenderedImpactSchema.renderer candidate, extracted
The renderer, by its Tool id or script, e.g.
fhir-ig-ast-pages.- RenderedImpactSchema.site candidate, extracted
The site the paths are relative to, as a reader finds it (an instance, a URL path).
- RenderedMeasuredSchema.baseCommit candidate, extracted
The commit the PR's change is measured against.
- RenderedMeasuredSchema.beforeCommit candidate, extracted
The commit the before side was built from, as its publish manifest records it.
- RendererContribution.adapters candidate, extracted
Which content adapters this target applies to.
- RendererContribution.format candidate, extracted
The output format this renders to —
latex,markdown,html. One contributor per format; two renderers forlatexis a collision, not a fallback chain.- RendererContribution.render candidate, extracted
Render one block's manifest plus its Markdown body to the target format.
- RendererContribution.validate candidate, extracted
Structural check on rendered output.
undefinedmeans this target has no structural check — reported as "not checked", never as a pass.- RenderExemption.of candidate, extracted
Which obligations are excused. Non-empty.
- RenderExemption.owes candidate, extracted
What it carries INSTEAD — the criterion it cannot fail quietly.
- RenderExemption.reachableAt candidate, extracted
Where the instance IS reachable, since it renders nothing of its own.
- RenderExemption.reason candidate, extracted
Why this layer is the exception. Prose, and required.
- RenderLogEntrySchema.branch candidate, extracted
Branch the render came from.
- RenderLogEntrySchema.capture candidate, extracted
Whether this entry is kept. Never defaulted silently.
- RenderLogEntrySchema.commit candidate, extracted
Commit the artefact was built from.
- RenderLogEntrySchema.format candidate, extracted
The declared format of
summary,detailandreason— see {@link RENDER_TEXT_FORMATS}.- RenderLogEntrySchema.id candidate, extracted
Stable id, so one entry can be pointed at from elsewhere.
- RenderLogEntrySchema.process candidate, extracted
BPMN process element id —
Process_RenderLog— when written inside one.- RenderLogEntrySchema.reason candidate, extracted
WHY, for the events where a reason is the whole content.
- RenderLogEntrySchema.references candidate, extracted
Everything this entry points at: a PR, an issue, a commit, a run.
- RenderLogEntrySchema.run candidate, extracted
The workflow run that wrote this, so a reader can open the log.
- RenderLogEntrySchema.summary candidate, extracted
One line, so a long log stays skimmable. Detail goes in
detail.- RenderLogEntrySchema.task candidate, extracted
The activity within that process.
- RenderOptions.compactInlineRefs candidate, extracted
In compact mode, include example/remark blocks that are directly referenced via
examples[]on definitions/theorems in the same view. Default: true. Set to false for a minimal compact output.- RenderOptions.printMode candidate, extracted
Print mode (default: "compact").
- RenderOptions.referencedTerms candidate, extracted
Document-wide set of
term:<slug>labels that some block refers to via:refterm[…]{#slug}/\refterm{slug}.- RenderSubjectSchema.path candidate, extracted
The path on the publish branch —
STAGING/<slug>, or/for the site.- RenderSubjectSchema.slug candidate, extracted
The staging slug, where the subject has one.
- RenderTargetDeclaration.argStyle candidate, extracted
How that export takes its two inputs.
- RenderTargetDeclaration.format candidate, extracted
Output format —
latex,markdown.- RenderTargetDeclaration.module candidate, extracted
Repo-relative module holding the renderer. Resolved by variable path.
- RenderTargetDeclaration.renderExport candidate, extracted
Export rendering one block manifest plus its Markdown body to
format.- RenderTargetDeclaration.validateExport candidate, extracted
Export checking rendered output is structurally sound, if there is one.
- RepositoryDescription.disagreements candidate, extracted
Facts two or more markers state differently.
- RepositoryDescription.types candidate, extracted
Every type whose marker is present, in registration order.
- RepresentationSchema.localPath candidate, extracted
Instance-relative path, present iff these bytes were materialised.
- Reproduction.both candidate, extracted
Both halves reproduced. Never true when either hash is
unknown.- Requirement.actors candidate, extracted
Actor IDs this requirement applies to.
- Requirement.derivedFrom candidate, extracted
IDs of requirements this one derives from.
- Requirement.description candidate, extracted
What this requirement governs.
- Requirement.id candidate, extracted
Unique requirement identifier.
- Requirement.statements candidate, extracted
Testable statements within this requirement.
Searchable tags.
- Requirement.title candidate, extracted
Human-readable title.
- RequirementFields.derivedFrom candidate, extracted
Broader requirements this one specialises, by id.
- RequirementFields.id candidate, extracted
req:<slug>.- RequirementFields.proposedIn candidate, extracted
WHERE IT CAME FROM — the proposal it was filed from, as a path, when it began as one. Kept after the proposal is moved, so the history can be followed back through
git log --follow.- RequirementFields.supersededBy candidate, extracted
When
superseded: what replaced it.- RequirementSetFields.document candidate, extracted
Where the document is, as a path or a link.
- RequirementSetFields.issue candidate, extracted
The issue (or other review surface) the set is decided on.
- RequirementSetFields.methodology candidate, extracted
The requirements methodology that governs it, by the harness's own name for it.
- RequirementSetFields.workPlan candidate, extracted
The work plan: work-item ids, one or more per member.
- RequirementStatement.actors candidate, extracted
Actors this statement applies to (defaults to parent's actors).
- RequirementStatement.conformance candidate, extracted
FHIR conformance verb.
- RequirementStatement.dependsOn candidate, extracted
Keys of other statements this one depends on.
- RequirementStatement.key candidate, extracted
Unique key within this requirement (e.g., "REQ-SS-1").
- RequirementStatement.label candidate, extracted
Short human-readable label.
- RequirementStatement.requirement candidate, extracted
Full requirement text.
- RequirementStatementFields.activity candidate, extracted
FUNCTIONAL — the activity the statement is part of.
- RequirementStatementFields.actors candidate, extracted
The roles it is about, by id in the harness's own role registry.
- RequirementStatementFields.benefit candidate, extracted
FUNCTIONAL — why ("so that …").
- RequirementStatementFields.capability candidate, extracted
FUNCTIONAL — what the actor wants to be able to do ("I want …").
- RequirementStatementFields.category candidate, extracted
NON-FUNCTIONAL — which quality it constrains.
- RequirementStatementFields.conformance candidate, extracted
How strongly it binds. Named
conformance— the word the harness's requirement files have used since they were first written — so that building them on this base moved no file.- RequirementStatementFields.dependsOn candidate, extracted
Other statements this one presupposes, by {@link requirementRef}.
- RequirementStatementFields.key candidate, extracted
Unique within its requirement; with the requirement's id it is the statement's address.
- RequirementStatementFields.requirement candidate, extracted
The statement itself, as one sentence a reviewer can say yes or no to.
- RequirementStatementFields.successCriteria candidate, extracted
How it will be judged met — at least one {@link SuccessCriterionSchema} when present. OPTIONAL FOR NOW, and deliberately so: statements filed before this field existed carry none, and they gain criteria by migration rather than by a grandfather date. A harness's own checker warns on a statement without it; once every filed statement carries one, the field becomes required.
- ResolvedBackdrop.art candidate, extracted
The art per layout. Complete, or empty — never partial; see below.
- ResolvedBackdrop.missing candidate, extracted
The layouts the instance does not supply for this theme's role.
- ResolvedBackdrop.none candidate, extracted
truewhen the theme declares no backdrop at all — not a gap.- ResolvedDependency.config candidate, extracted
The dependency's own harness config (if present).
- ResolvedDependency.dependency candidate, extracted
The original dependency declaration.
- ResolvedDependency.needs candidate, extracted
Absolute roots of this dependency's OWN dependencies. Edges, not a subtree: a diamond's shared node is one object reached twice, and a nested copy of it under each branch was how the old walk lost it.
- ResolvedDependency.rootPath candidate, extracted
Absolute path to the dependency root.
- ResolvedDirectory.absPath candidate, extracted
Absolute path, resolved against the instance that declared it.
- ResolvedDirectory.declaredBy candidate, extracted
Name of the instance whose declaration contributed this entry.
- ResolvedDirectory.member candidate, extracted
The instance whose directory this is. A declared subgraph is inherited as one MEMBER per instance in the chain whose same-named directory exists (owner, 2026-09-30, bean
3r47/xsqmfollow-up, option A: "if f-a-core … has a docs/ … and depends on cat-harness where docs/ is declared … the viewer should detect f-a-core/docs", "a named (sub-)subgraph of docs"). The declarer's own member comes first;membernames each.- ResolvedDirectory.own candidate, extracted
True when this entry is the ROOT instance's — its own declaration, or its own member of an inherited subgraph.
- ResolvedDirectory.within candidate, extracted
The id of the directory that declared this one FROM WITHIN (bean
cmsl, owner 2026-09-30): an entry of its kind's declaration file, e.g.skills/skills.jsonnamingvoices/. Absent for an instance-level entry.check:layout-normsreads it: nesting declared this way is the sanctioned shape, not a root declaration reaching down.- ResolvedTag.axis candidate, extracted
Which axis —
role,process,task,actor.- ResolvedTag.ref candidate, extracted
The reference as written, e.g.
crdm-requirements▸A_Implement.- ResolvedVoice.appliesTo candidate, extracted
The START voice's own declaration, kept for a consumer reporting on the voice as authored. NOT the effective scope of its resolved rules — that is {@link ResolvedVoice.scopeOf}, and conflating them is the defect above.
- ResolvedVoice.chain candidate, extracted
Root-first, so a reader can see where each rule entered.
- ResolvedVoice.origin candidate, extracted
Which voice in the chain contributed each rule, by rule id.
- ResolvedVoice.scopeOf candidate, extracted
The scope each rule was DECLARED under, by rule id — not the resolved voice's own.
- ResolvedZoom.because candidate, extracted
Present exactly when
sourceis"kind"— an override always cites why.- ReviewCommentsFileSchema.commit candidate, extracted
The head commit whose blocks the comments were anchored against.
- ReviewCommentsFileSchema.verdicts candidate, extracted
Reviewers' per-block verdicts (
folio-review-verdict/v1, beanpx0t), ingested from the same PR comments. Default[], so a file written before verdicts existed still reads.- ReviewerRef.id candidate, extracted
Script path, agent name + skill, or GitHub login.
- ReviewerRef.version candidate, extracted
Script revision or model id.
- ReviewerSchema.id candidate, extracted
Stable pseudonym: a hash of the normalised email or name.
- ReviewerSchema.name candidate, extracted
Only when
acknowledgeis true.- ReviewFieldsSchema.anchoredFrom candidate, extracted
The labels this comment was anchored to before a rename re-anchored it.
- ReviewFieldsSchema.blockHash candidate, extracted
The block's content hash when the comment was ingested.
nullwhen the block was not in the head at all, so the comment was orphaned from the start. Recorded rather than refused, because a reviewer's words are kept.- ReviewFieldsSchema.commentId candidate, extracted
GitHub's id of the conversation comment this was ingested from.
- ReviewFieldsSchema.commit candidate, extracted
The head commit the comment was ingested against.
- ReviewFieldsSchema.decision candidate, extracted
The 9gyz Decision that closed it. Required once resolved or adjudicated.
- ReviewFieldsSchema.decisions candidate, extracted
The PR conversation comments whose
resolve:tag has moved this comment ({@link applyDecisions}), in the order applied. A rebuild applies each one once: the record, not the build count, says what has been decided.- ReviewFieldsSchema.orphaned candidate, extracted
The block the label names is gone. Kept and shown, never dropped. An anchor fact, not a status: the review process still decides what happens to the comment.
- ReviewFieldsSchema.pr candidate, extracted
The edit-set's pull request.
- ReviewFieldsSchema.repo candidate, extracted
owner/nameof the repository the pull request is in.- ReviewFieldsSchema.reviewer candidate, extracted
GitHub login of the reviewer.
- ReviewFieldsSchema.role candidate, extracted
The lane they reviewed in: a role id from
roles.json.- ReviewProblem.where candidate, extracted
The offending node — a finding id, a decision id,
review.- ReviewRead.meta candidate, extracted
The legacy file's
_meta, when that is where the read came from.- ReviewRead.reviews candidate, extracted
Reference id → review, in recorded order.
- ReviewRecord.confidence candidate, extracted
Reviewer's confidence in the verdict (0–1).
- ReviewRecord.reviewer_id candidate, extracted
Identifier for the reviewer (model name, GitHub username, etc.).
- ReviewRecord.timestamp candidate, extracted
ISO-8601 timestamp.
- ReviewTag.text candidate, extracted
Everything after the header, trimmed.
- ReviewTransition.from candidate, extracted
nullfor creation.- ReviewTransition.needsDecision candidate, extracted
A transition that closes a comment must name the Decision that closed it.
- ReviewVerdictSchema.blockHash candidate, extracted
The block's content hash (a page's or input's pin) when the verdict was given. It counts only while current.
- ReviewVerdictSchema.commit candidate, extracted
The head commit the verdict was ingested against.
- ReviewVerdictSchema.id candidate, extracted
verdict-pr<n>-c<comment id>-<label>: derived, so a re-run finds it rather than adding one.- ReviewVerdictSchema.reason candidate, extracted
Required for
waived: why this block needs no review.- ReviewVerdictSchema.reviewer candidate, extracted
GitHub login.
- ReviewVerdictSchema.role candidate, extracted
The lane they reviewed in: a role id from
roles.json.- ReviewVerdictSchema.target candidate, extracted
Absent on verdicts written before pages could be reviewed: those are all on blocks.
- ReviewVerdictSchema.targetLabel candidate, extracted
A block's label, or a page's or an input's path, as
targetsays.- RoleAssignment.actorId candidate, extracted
Actor ID to assign.
- RoleAssignment.identitySource candidate, extracted
How the user's identity was determined.
- RoleAssignment.priority candidate, extracted
Higher priority wins.
- RoleAssignment.userPattern candidate, extracted
Glob pattern matched against the user identifier.
- RoleDef.actorKinds candidate, extracted
Which kinds of actor may take this role on.
externalmarks a participant outside the instance's control (a registry, a third-party service).- RoleDef.formerNames candidate, extracted
Names this role was known by and no longer is, each with the date it was retired —
InitiatorbecameBootstrapping Agenton 2026-09-23. Published asskos:hiddenLabelon this role, and the old name's concept isowl:deprecatedanddcterms:isReplacedBythis role, so a reader holding the old name is sent to the new one. Authored here, beside the definition, rather than only remembered by the harness's glossary ledger: a retired name is part of what the role IS to a reader of old text.- RoleDef.id candidate, extracted
Stable id. It is the
idof the<bpmn:resource>a lane'spartitionElementRefpoints at, verbatim, and whatinheritsnames.- RoleDef.inherits candidate, extracted
Roles this one IS-A. Skills are unioned transitively; cycles rejected.
- RoleDef.judgementOnly candidate, extracted
The role PERFORMS, but by judgement — no instruction body implements its steps, and naming one would be a lie about what the role does.
- RoleDef.otherNames candidate, extracted
Other names this role goes by today — synonyms, local titles, examples. The same field, and the same meaning, as
otherNameson WHO SMART Base's Generic Persona ("Other names or examples for the persona"); published asskos:altLabel. Owner, 2026-09-30: *"model both retired names and alternative names"*.- RoleDef.persona candidate, extracted
Who this reader IS, in prose — the persona an author writes for and a QA reviewer checks against.
- RoleDef.requiredCapabilities candidate, extracted
Minimum model capabilities required by this role when performed by an agent/model. If a single model lacks these capabilities, the workflow refuses rather than degrades.
- RoleDef.skills candidate, extracted
Skills available to an actor in this role, before inheritance.
- RoleDef.title candidate, extracted
Display text, and the
nameof that<bpmn:resource>. Not used for matching: a lane binds by id.- RoleDefSchema.formerNames candidate, extracted
Names it no longer goes by, with the retirement date;
skos:hiddenLabel.- RoleDefSchema.inherits candidate, extracted
Roles this one IS-A, outermost last. Static composition, not the scoped subprocess stack.
- RoleDefSchema.otherNames candidate, extracted
Other names this role goes by today;
skos:altLabel. As smart-base'sotherNames.- RoleDefSchema.skills candidate, extracted
Skills available to an actor in this role, before inheritance.
- RoleExtension.role candidate, extracted
The id of a role declared by an instance this one depends on.
- RoleExtension.skills candidate, extracted
Skills the role gains in this instance's checkout.
- RoleGraph.actors candidate, extracted
Actors, when declared alongside. Usually read from the actor directory.
- RoleGraph.extensions candidate, extracted
Roles of a LOWER instance this graph adds skills to, BY ID (placement PR0b, bean
ejye). See {@link RoleExtension}.- RoleStack.path candidate, extracted
Role ids, outermost first. Unresolvable ids are kept and flagged.
- RoleStack.unresolved candidate, extracted
Ids in
paththat are not declared.- RouteClaim.by candidate, extracted
Who claims it, for the message:
cat-harness.visualisers[library],alias of …,site: processes/.- RouteClaim.route candidate, extracted
Site-relative route, trailing slash for a directory.
- RouteCollision.kind candidate, extracted
same— one route claimed twice;nested— a claim inside another's subtree.- RouteManifestSchema.gate candidate, extracted
The command that checks the generator's output.
- RouteManifestSchema.writer candidate, extracted
The ONE generator that writes it.
- RunContextSchema.profile candidate, extracted
The tool profile selected. Absent: the run selected none, and
releases[]lists what it resolved anyway.- RunContextSchema.prov candidate, extracted
The PROV activity that records this context as
prov:used. Absent until the run records one (step 3).- RunInvokerSchema.id candidate, extracted
The script path, the model id, the person's handle.
- RunInvokerSchema.script_hash candidate, extracted
Hash of the script's bytes, or
unknown. Required for asysteminvoker.- RunRecord.blockedBy candidate, extracted
Why it was skipped: the id whose failure reached it.