Project: Covert Cubicles Build Phase 6 / 7 · 2 of 6 modules complete —

MODULE 21 / 31 · LAYER 6 · DATA_RECORDER · EXPORT Per-Session Record Builder

Bundles M20 telemetry plus the seven SM_Game logs and the audit trail into a canonical session record. Provides JSON serialisation and three opinionated CSV views for the most-needed analytical use cases. Standardised filename schema for save-to-disk workflows.

DEPENDENCIES: SM_Game, DATA_Telemetry

Pending

DRAFT EXPORT POSTURE \u00B7 ANONYMISATION PENDING REVIEW

Default export is "as-is" — sessionId is already a UUID per SM_Game's de-identification commitment.

Per ADR-059, the current export anonymisation is "as_is" — sessionId is the SM_Game UUID and no PII is captured anywhere in the session record. This is a reasonable default for a pre-pilot build but warrants explicit supervisor and ethics-committee review before participant data collection. Stricter postures (strip-and-link-table, configurable per-export) are not yet implemented; the module exposes a stub setter for future expansion.

Latent integration risk \u2014 dual baseline-values store (ADR-061)

Two separate baseline-values stores exist: SM_Game.playerState.ethicalValues (canonical, what M21 reads) and ALMAA_Morality's internal preGameValues (computing copy). Field-name conventions differ (pa vs pa_score). Not synchronised. M31 GM_Game integration must call both setters consistently — preferably by having M17 read from SM_Game's canonical store. Recorded as defect for resolution at integration time, not patched now.

Automated Assertions

#	Assertion	Result	Detail

Interactive Workbench — Live Session Record

Drive the session lifecycle on the left; the canonical session record updates after each action. Log cards in the viewer flash amber when their counts change. The export tabs below show JSON and CSV views generated on demand.

Session lifecycle harness

Session creation + baseline

Phase 1 decisions (build pipeline activity)

Scrutiny escalation

Final morality record (RQ1 PRIMARY DV)

End session

Reset

Live session record \u2014 buildSessionRecord()

recordVersion—

sessionIdnull

conditionVariantnull

terminationReasonno_session

decisions0 / 16

filename—

baselineValuesnull

                    LOGS \u2014 8 ARRAYS, COUNTS + LATEST ENTRY PREVIEW
                

Export views \u2014 JSON + 3 CSV views

{}

size — bytes

Record schema v1.0 \u2014 ADR-057

Field	Type	Description
Provenance
recordVersion	string	Schema version of THIS record (e.g. '1.0')
recordedAt	ISO 8601 UTC	When the record was built
telemetryVersion	string	Schema version of nested telemetry blob
Nested telemetry (M20 schema)
telemetry	object	Full M20 telemetry blob — see DATA_Telemetry schema
Logs (verbatim from SM_Game)
logs.decisionHistory	array	Decisions submitted
logs.consequenceTags	array	Story-flag effects from M11
logs.events	array	Narrative events fired by M14
logs.confusion	array	M15 confusion classifications
logs.reasoning	array	M16 reasoning classifications
logs.morality	array	M17 morality records (PRIMARY DV)
logs.trigger	array	M18 trigger decisions
logs.condition	array	M19 condition gating decisions (RQ1b)
Audit + baseline
auditTrail	array	SM_Game lifecycle events in chronological order
baselineValues	object \| null	Pre-game survey values { pa, os, ei } if set
CSV views (separate methods, not in record)
toDecisionsCsv()	string	Flat tabular: decisionId, choiceId, phase, timestamp, scrutinyLevel
toTriggerLogCsv()	string	Flat tabular: trigger records with intended/blocked/precedence
toMoralityRecordsCsv()	string	Flat tabular: morality records (RQ1 substrate)

API Surface — Manual Verification

M21 has a more substantial API than M20 — JSON, CSV, filename, and export options.