O8A Report 02 — v0.5 proven-module mapping

• macro: v0.6-o8a-live-execution-wiring-authoring
• date_utc: 2026-05-21 · host: vmi3080463 (Contabo)
• gate covered: G1 v0.5 proven-module mapping

1. Survey of the v0.5 write modules in the v0.6 tree

cutter_agent/ ships the v0.5-ratified write modules. Surveyed for re-use as the orchestrator's Mode.LIVE backing.

Decisive finding — the v0.5 modules are Constitution-cut scripts

They are execution scripts pinned to the 60-IU Constitution, not generic libraries:

• prod_iu_adapter_canonical.py — module-level PIN_WRITER_DIGEST, PIN_CANDIDATE_COUNT = 60, G6 check literally WHERE canonical_address LIKE 'ICX-CONST%'.
• ledger_v2_canonical_cut.LegBRecorder — .plan() calls _require_list(live_state, k, PIN_CANDIDATE_COUNT); hard-pinned to N=60. PIN_CUT_COMMITTED_AT is a fixed timestamp.
• ledger_v2_canonical_verify.VerifyRecorder — _require_int / _require_uuid against Constitution pins.

➡️ They cannot be reused as drop-in N-agnostic callables. They can be reused as the proven SQL contract + transaction doctrine layer. Re-using them blindly would fail-closed for any non-60 document (safe) but is not honest "wiring". O8A therefore re-uses the proven SQL/txn patterns and isolates the N-coupled classes behind injectable writer seams.

2. Per-phase mapping

Phase	v0.5 proven source	Re-use in O8A
pre_write_backup	no Python module — pg_dump --table + GPG is an operational procedure (v0.5 G7 backup gate was a backup_gate callable)	seam: injected backup_runner(spec); adapter pins sha/size/fpr
cut_leg_a	prod_iu_adapter_canonical.leg_a_in_txn_canonical — SELECT public.fn_iu_create(%s×9) per row, status='created', one caller-owned txn	SQL contract reused in full, parameterised per-run (generic)
leg_b_record	ledger_v2_canonical_cut.LegBRecorder.record(conn, live_state) — 8-table governed insert, G-LEG-B-ONCE	seam: adapter owns the txn, delegates writes to an injected governance_writer (production = wraps LegBRecorder)
write_verify	ledger_v2_canonical_verify.VerifyRecorder.record(conn, live_state) — cutter_verify lane, G-VERIFY-ONCE	seam: adapter owns the txn, delegates to injected verify_writer (production = wraps VerifyRecorder)
lifecycle_enact	no Python module — public.fn_iu_enact(text,text,uuid,text,uuid,text,text,boolean) (SQL fn, sql/lifecycle/bundle_c_fn_iu_enact.sql)	SQL contract authored fresh, same per-row fan-out doctrine as cut_leg_a

3. fn_iu_enact signature (from bundle_c_fn_iu_enact.sql)

public.fn_iu_enact(
  p_canonical_address text, p_actor text, p_review_decision_id uuid,
  p_reason text, p_change_set_id uuid DEFAULT NULL,
  p_tool_revision text, p_<aux> text, p_dry_run boolean
) RETURNS jsonb           -- success status = 'enacted'

4. Mapping verdict

G1_v0_5_module_mapping: PASS
reusable_as_is:        none (all v0.5 write modules are Constitution-N-pinned)
reusable_as_contract:  fn_iu_create SQL (cut_leg_a), fn_iu_enact SQL
                       (lifecycle_enact), LegBRecorder/VerifyRecorder
                       (behind an injected writer seam), txn doctrine
                       (autocommit=False; caller-owned BEGIN/COMMIT/ROLLBACK;
                        atomic all-or-nothing — v0.5 G6 lesson)
authored_fresh:        pre_write_backup runner seam; fn_iu_enact fan-out
follow_up_gap:         LegBRecorder/VerifyRecorder N=60 generalisation —
                       isolated behind the governance_writer/verify_writer
                       seam; out of the O8A 5-phase wiring scope.

No mapping was impossible → G1 = PASS (not a BLOCKED gap).