diff --git a/docs/superpowers/plans/2026-06-20-cyclone-edi-features.md b/docs/superpowers/plans/2026-06-20-cyclone-edi-features.md
index e466a52..ec34b39 100644
--- a/docs/superpowers/plans/2026-06-20-cyclone-edi-features.md
+++ b/docs/superpowers/plans/2026-06-20-cyclone-edi-features.md
@@ -56,8 +56,15 @@ def test_r034_ref_g1_required_when_freq_1_no_ref_g1_passes(strict_ref_g1_payer):
 
 **Files:**
 - Modify: `backend/src/cyclone/parsers/validator.py`
-- Modify: `backend/src/cyclone/parsers/models.py` (add `Envelope.transaction_type_code: str | None`)
-- Modify: `backend/src/cyclone/parsers/parse_837.py` (populate `transaction_type_code` from BHT06 in `_build_envelope`)
+- Modify: `backend/src/cyclone/parsers/models.py` (add `ClaimOutput.transaction_type_code: str | None = None`)
+- Modify: `backend/src/cyclone/parsers/parse_837.py` (populate `ClaimOutput.transaction_type_code` from BHT06 in the per-claim builder)
+
+> **Why on ClaimOutput, not Envelope?** The current `ClaimOutput` shape
+> already mirrors a few envelope fields (`control_number`,
+> `transaction_date`) because the per-claim flow is easier when the
+> validator only needs to look at one object. Following the same
+> convention keeps R035's lookup one-liner clean and avoids threading
+> the Envelope into every validation rule signature.
 
 > **Note on naming:** BHT06 in the 837P 005010X222A1 implementation is the
 > "Transaction Type Code" element (qualifier 640) — not the "Transaction