Generated from cat-harness/sdlc/processes/test-data-curation.bpmn by gen-processes-viz.ts — do not edit here. All processes · cat-harness processes

Test Data Curation — Fixed Sets and Generated Recipes

Process_TestDataCuration · strict · 9 step(s)

General curation process for fixed (test-data-fixed/1.0.0) and generated (test-data-generated/1.0.0) test data sets. It opens its bean with Process_BeanOpen before any work and closes it with Process_BeanClose before it ends; the steps between note on it.

SEPARATE GRAPH DIRECTORIES, SELF-IDENTIFYING FILES. Fixed sets live in a declared test-data-fixed directory and generated recipes live in a declared test-data-generated directory (schemas/kg/graph-typology-registry.ts). Every file carries both its $schema tag (test-data-fixed/1.0.0 vs test-data-generated/1.0.0) and kind: "fixed" | "generated" (schemas/sdlc/test-data.ts), so a consumer (discriminateTestDataSet) tells which it is reading from the file content, never from the directory path.

CONTENT-TYPE SPECIALISATION AS A CALLED SUBPROCESS. How a recipe synthesises test inputs or how a fixed payload is structured depends on the folio’s contentType. Call_ContentTypeSpecialisation delegates that step to the content type’s own subprocess (specialisation in schemas/sdlc/test-data.ts). FHIR templating (specialisation: "fhir-templating") is one named specialisation for FHIR IG / DAK content, not the platform design.

BPMN diagram: Test Data Curation — Fixed Sets and Generated Recipes

How it connects

Lanes — who acts

lanerolewhat it does here
Authoring agentauthoring-agentAuthors or curates the test data set: classifies whether the case requires a fixed curated set or a deterministic generated recipe, invokes the content-type specialisation subprocess on the generated branch or SME review on the fixed branch, and writes the self-identifying node into its declared graph directory.
TestertesterVerifies that the curated fixed set or generated recipe self-identifies from its file content (discriminateTestDataSet) rather than from its directory path, verifies the canonical SHA-256 digest, and binds the TestDataRef into the test-plan/v1 test case.

Steps

Every one of the 9 step(s) is documented.

steplaneskill / sub-processwhat it does
Classify data set: fixed (curated) or generated (recipe)
Task_ClassifySet
Authoring agenttest-data-setsDecide whether the test case requires a small, committed, cited fixed data set (test-data-fixed/1.0.0, usable as certification evidence once reviewed) or a deterministic generated recipe (test-data-generated/1.0.0, template + params + seed, never stored materialised).
Author the fixed test data set
Task_CurateFixed
Authoring agenttest-data-setsWrite the curated records and optional citations with $schema: "test-data-fixed/1.0.0" and kind: "fixed" into the layer’s declared test-data-fixed directory (schemas/sdlc/test-data.ts). Recipe fields (template, params, seed) are refused by the strict schema.
SME review of cases or cohorts
Call_SmeReview
Authoring agentcalls SME Review of Decision Support and Indicator Test Data
test-data-sets
Calls Process_SmeTestDataReview (sdlc/processes/sme-test-data-review.bpmn) so a human SME in a judgementOnly lane reviews each worked case (decision-support) or cohort plus expected value (indicator-definition) and records an attributable sme-case-verdict/1.0.0 node.
Determine the content type and its specialisation
Task_DetermineSpecialisation
Authoring agenttest-data-setsRead the folio’s contentType and the specialisation registered for it (TestDataSpecialisationSchema in schemas/sdlc/test-data.ts): fhir-templating for FHIR IG and DAK content, markdown-fixture for document folios, lean-counterexample for proof folios, tabular-cohort for aggregate indicator tables. None registered stops here and says so, rather than guessing one.
Synthesise the recipe with that specialisation
Call_ContentTypeSpecialisation
Authoring agenttest-data-setsBOUND BY THE CONTENT TYPE, so no calledElement is written here: each specialisation process belongs to the layer that owns its content type, and a lower layer cannot name a higher layer\’s process. The specialisation is the one Task_DetermineSpecialisation found; so no single content-type calledElement is hardcoded here: the activity names test-data-sets, and each content-type adapter supplies its own specialisation subprocess. fhir-templating (FHIR_TEMPLATING_SPECIALISATION) is one named specialisation for FHIR IG / DAK content (MeasureReport cohorts, questionnaire responses, bundle templates) — named as such, not the platform design; document folios (markdown-fixture), proof folios (lean-counterexample), and tabular cohorts (tabular-cohort) supply their own.
Write the generated-data recipe
Task_WriteRecipe
Authoring agenttest-data-setsRecord contentType, specialisation, template, params, seed, and the canonical sha256 digest (computeGeneratedSetDigest) with $schema: "test-data-generated/1.0.0" and kind: "generated" in the layer’s declared test-data-generated directory. Never store materialised records.
Verify file self-identifies by $schema and kind and bind into test plan
Task_VerifySelfIdentifying
Testertest-data-setsParse the file with discriminateTestDataSet (schemas/sdlc/test-data.ts) to confirm a consumer can tell fixed from generated from the file content alone (never the path), verify verifyGeneratedSetDigest on generated sets, and bind the resulting TestDataRef (toTestDataRef) into the test-plan/v1 case.
Open the bean
Call_BeanOpen
Authoring agentcalls Open the bean
bean-coordination
Process_BeanOpen: check before you create, find out whose the bean is, and claim it — or leave it alone when it is someone else’s. The one place a claim is drawn; the gateway after this reads which end it reached.
Close the bean
Call_BeanClose
Testercalls Close the bean
todo-manager
Process_BeanClose: end the bean in a state that says why — complete and archive it, scrap it with its reasons and archive it, or record the blocker and hand it back. The one place resolve and archive are drawn; placed straight before the end so the engine’s resolve sees nothing left but the archive step.

Decisions

Every one of the 3 decision(s) is documented.

decisionwhat decides itbranches
Fixed or generated?
GW_SetKind
Routes to the fixed-set curation branch (fixed) or the generated-recipe synthesis branch (generated).fixed → Author the fixed test data set
generated → Determine the content type and its specialisation
Exercises decision support or indicator definition?
GW_NeedsSmeReview
When the fixed data set exercises clinical/scientific decision support or an indicator definition, it must pass SME review (Process_SmeTestDataReview) before review is recorded and the set can stand as certification evidence.yes → SME review of cases or cohorts
no → Fixed set ready
Bean claimed?
GW_BeanYours
Reads which end Process_BeanOpen reached. claimed: the bean is in progress and this process works it. not yours: a sibling session or a human holds it, nothing was written to it, and this process stops here rather than work a bean it does not own (bean-coordination).claimed → Classify data set: fixed (curated) or generated (recipe)
not yours → Not yours — left alone

folio-assistant — code under the Apache License 2.0, documentation under CC BY 3.0. Generated schema and API references are rebuilt from source on every push to main.

deployed dev