Status: Adopted (agent-recorded design decision) — pending named
architecture ratification Date: 2026-07-24 Source of truth for the
classification data:
evidence/v1-workbenches/component-extraction-matrix.json (regenerated +
coherence-checked by
node scripts/v1-workbenches/generate-component-extraction-matrix.mjs)
Grounded in: evidence/v1-workbenches/tara-component-inventory.json
(§S0.2.a/b — 29 components with per-component
props/state/data/mutation/token/test columns and a meditation-vocabulary signal)
This document records the §S0.2 classification decisions that the deterministic inventory (a/b) cannot make on its own:
- S0.2.c — behavior independently usable by ≥1 non-Tara archetype (neutral candidates), with rationale.
- S0.2.d — meditation vocabulary/semantics that must remain Tara-owned, with rationale.
- S0.2.e — mixed components split at an explicit seam or rejected, per row.
- S0.2.f — neutral candidates validated against an authoring and an operations-console consumer (two-archetype sketches).
The extraction seam, in every case, is the interaction contract (a comment thread, a review-decision, a gate run, a scheduling calendar…), parameterized over a domain-supplied data adapter and vocabulary — never the wired component, whose Tara routes and copy stay Tara-owned. Per §S0.11 (deletion threshold) no extracted contract leaves Tara until a real 2nd consumer exists; the two-archetype sketches identify who that consumer would be.
Summary (29 components)#
| Bucket | Count | Meaning |
|---|---|---|
| Neutral candidate | 4 | 0 meditation-vocabulary terms; universal workbench pattern; extract the contract as-is |
| Tara-owned | 13 | Meditation semantics in the contract/copy/data; must remain in Tara |
| Mixed → split | 9 | Neutral pattern under Tara vocabulary; extract the shell, keep Tara wiring |
| Mixed → reject | 2 | Neutral residue too thin to justify a shared contract; keep whole in Tara |
| Already-shared | 1 | App-shell infrastructure (ShellLayout), out of kit scope |
S0.2.c — Domain-neutral candidates#
The 4 zero-signal components plus the neutral contract yielded by each mixed-split row (below) are the domain-neutral surface. Each is independently usable by a non-Tara archetype (see S0.2.f sketches).
| Component | Neutral contract | Why neutral |
|---|---|---|
StudioTaraWorkbenchAssignments |
AssignmentPanel |
assign/clear an actor against an object under review — universal |
StudioTaraWorkbenchComments |
CommentThread |
threaded, resolvable comments — universal |
StudioTaraWorkbenchReview |
ReviewDecisionPanel |
record a capability-gated decision against an object — universal |
StudioTaraWorkbenchFunnel |
PipelineFunnel |
stage-count funnel visualization — pure presentation |
S0.2.d — Tara-owned (must remain)#
These carry meditation vocabulary/semantics in their contract, copy, or data shape; extraction would leak Tara semantics into a "neutral" package.
| Component | Signal terms | Why Tara-owned |
|---|---|---|
StudioTaraWorkbenchDevelop |
invitational/breath/grounding | meditation-script compile/readability craft |
StudioTaraWorkbenchProgramDetail |
invitational/steering/seasonal/spark/cadence | meditation-program cadence/seasonal planning (highest signal) |
StudioTaraWorkbenchPrograms |
steering/seasonal/cadence | meditation-program planning list |
StudioTaraWorkbenchMeditationCuration |
meditation | curation of meditation catalog (name + kind filter) |
StudioTaraWorkbenchCollectionEditor |
meditation/seasonal | Tara collection semantics |
StudioTaraWorkbenchConceptBoard |
spark/maturity | Tara ideation (a "spark" is a Tara unit) |
StudioTaraWorkbenchInbox |
breath/spark | sparks?status=inbox ideation inbox |
StudioTaraWorkbenchDivergence |
divergence/spark | Tara's creative divergence method |
StudioTaraWorkbenchOverview |
spark/plane | Tara dashboard composition |
StudioTaraWorkbenchWorkspace |
meditation/grounding | Tara workspace composition |
StudioTaraWorkbenchSteering |
steering | Tara agent-direction vocabulary |
StudioTaraWorkbenchDossier |
grounding/spark | Tara concept dossier (thin generic residue) |
StudioTaraWorkbenchPlaneLinkage |
plane | links to a Tara "plane" taxonomy node |
S0.2.e — Mixed components: split at a seam, or reject#
Split (9) — extract the neutral contract; Tara supplies the adapter + vocabulary:
| Component | Neutral contract | Seam / what Tara keeps |
|---|---|---|
StudioTaraWorkbenchGrounding |
EvidenceGroundingPanel |
claim→citation→grounding-check contract; Tara keeps claim copy + source scoring (shared with Veritas) |
StudioTaraWorkbenchGates |
GateRunnerPanel |
run a declared gate set + render pass/fail; Tara supplies its gates |
StudioTaraWorkbenchCatalog |
CatalogBrowser |
filter/search/grid shell; Tara supplies row schema + kind facet |
StudioTaraWorkbenchCalendar |
SchedulingCalendar |
calendar + committed-slot reservation; Tara supplies cadence/plane events |
StudioTaraWorkbenchPublishing |
BundlePublisher |
assemble→attach→publish bundle; Tara supplies bundle schema |
StudioTaraWorkbenchSources |
SourceLibrary |
source list + bulk import; Tara supplies source schema (shared with Veritas) |
StudioTaraWorkbenchSourceDetail |
SourceInspector |
inspector + blast-radius impact view; Tara supplies rescore heuristic |
StudioTaraWorkbenchCategories |
TaxonomyEditor |
hierarchical taxonomy CRUD; Tara supplies the "plane" labeling |
StudioTaraWorkbenchAgentBudget |
AgentAutonomyBudget |
consumes the shared /v1/agentic/* surface already; Tara keeps only plane grouping |
Reject (2) — the neutral residue is too thin to justify a shared contract:
| Component | Why reject |
|---|---|
StudioTaraWorkbenchBriefLauncher |
brief + run-links launcher tightly bound to Tara concept/brief/run semantics; residue ("launch a job") too thin |
StudioTaraWorkbenchAgentConcepts |
no direct data request, plane-coupled grouping; not enough neutral surface — revisit if a 2nd agentic console needs it |
S0.2.f — Two-archetype usage sketches#
Each extractable neutral contract is validated against an authoring console
and an operations console (the two archetypes S0.7 models), demonstrating
reuse independent of Tara. Full text in component-extraction-matrix.json
(.archetypeSketches); representative rows:
| Contract | Authoring console | Operations console |
|---|---|---|
CommentThread |
comments on a draft chapter, resolved on edit | comments on an incident timeline, resolved on closure |
AssignmentPanel |
assign a reviewer to a draft asset | assign an operator to a processing job |
ReviewDecisionPanel |
approve/return a draft {approve, revise, reject} |
ack/resolve an alert {ack, mitigate, close} |
PipelineFunnel |
concept → review → publish funnel | job queued → running → done → failed funnel |
EvidenceGroundingPanel |
Veritas: ground article claims in cited sources | attach + cite evidence on an incident RCA |
GateRunnerPanel |
run pre-publish quality gates | run pre-deploy readiness checks |
SchedulingCalendar |
content-release calendar w/ reservable slots | maintenance-window calendar w/ reservable windows |
SourceLibrary |
Veritas research sources + bulk import | data-feed/integration sources + bulk import |
Consequences#
- The extraction target is a set of 13 neutral interaction contracts, not 13
wired components. Each becomes a
@oshun/workbench-kitexport only when a real 2nd consumer exists (§S0.11), consistent with ADR-S1 (staged, threshold-gated hybrid). EvidenceGroundingPanelandSourceLibraryare the strongest early candidates: Veritas already has structurally identical claim-grounding and source-management surfaces, so a real 2nd consumer is near at hand.- Named architecture + affected-owner ratification of this matrix remains a human governance step (tracked with the broader S0 signoff); the classification is coherence-gated in CI so it cannot silently drift.
References#
V1_DOMAIN_WORKBENCHES_TODOS_2026-07-23.md§S0.2evidence/v1-workbenches/tara-component-inventory.json(§S0.2.a/b)evidence/v1-workbenches/component-extraction-matrix.json(this matrix, machine-checked)docs/adr/ADR-S1-workbench-kit-package-ownership.md(§S0.6 ownership + dependency direction)evidence/v1-workbenches/deletion-threshold.json(§S0.11 ≥2-consumer gate)