Production Lines
WO-0001 — Edward-v4 parity gate result
WO-0001 — Edward-v4 parity gate result
Order: WO-0001 (STM EA Agent) Phase: Edward-v4 migration, Phase 4 (parity gate) Date: 2026-05-31 Result: PASS
What was tested
The v4 acceptance criterion: the talent-driven chain (Edward running /ea-hlsd)
produces the same deliverable as the engine-driven pipeline, with zero
per-client HTML hand-editing. Worked example: Transgesco DAE-0007 (the engine's
proven dataset — 273 catalog objects → 4 Macroscope pages).
Evidence
| Check | Result |
|---|---|
| Talent == engine (structural) | ea-hlsd v4 stages 6–8 delegate to export-datajs.mjs / publish.mjs / d1-export.mjs (Bash). The talent never hand-renders HTML — the deterministic engine is the only renderer. |
| Object/page counts | 273 objects → 4 pages (matches the Phase-B proven output) |
| Determinism | Two independent publish runs → all 9 output files byte-identical |
| Verbatim editorial layer | out/data.js === source curated data.js (byte-for-byte) |
| Zero hand-editing | Each generated HTML page === its token-filled template ({{CLIENT}}/{{DAE}}/{{HOST}} only) |
| Durable backup | Layer-1 catalog-data.js generated independent of output format |
Harness
- Ad-hoc Node parity check over
production-lines/agent-ea/pipeline/examples/transgesco-dae-0007.manifest.json. - Permanent regression:
production-lines/agent-ea/pipeline/publish.test.mjs(full pipeline suite 8/8 green). - Recorded in
production-lines/agent-ea/playbook/engagement.md(Acceptance proof).
Implication for handover
Phase 4 (the gate that had to pass before STM handover, decision 2A) is cleared.
Remaining before handover: Phase 5 (v4 deploy manifest/checklist + real STM manifest)
and Phase 7 (re-deploy the lean v2 onto the live OneDrive copy — the live-client
mutation, now unblocked by this gate).