O8E pre-production hardening (Contabo) — 06-safe-target-discovery
O8E Report 06 — Safe target discovery & acceptance criteria (G6 / GAP8)
- macro:
v0.6-o8e-pre-production-hardening-bundle - date_utc: 2026-05-21 · host: Contabo
vmi3080463 - gate covered: G6 — GAP8 safe target discovery
- result: G6 PASS — candidate space surveyed; acceptance criteria defined; intake package produced
1. Discovery — what is registered (read-only, public schema)
public.information_unit breakdown (158 IUs, by document root × lifecycle):
ICX-CONST: 60 units — ALL 'enacted' (the M1 Constitution cut — DONE)
test: 7 units — 'draft'
D38-DIEU28-*: ~26 units — 'draft' (1 unit each, ~26 distinct roots)
D38-DIEU32-*: ~24 units — 'draft' (1 unit each)
D38-DIEU35-*: ~36 units — 'draft' (1 unit each)
pilot.*: 5 units — 'draft'
TOTAL: 158 (60 enacted + 98 draft) — matches the O8D baseline
2. Discovery limitation — registry not readable by the RO role
The orchestrator cuts a source document version identified in
cutter_governance.source_document_registry /
source_document_version_registry. The read-only query_pg MCP role is
privilege-denied on the entire cutter_governance schema
(SQLSTATE 42501 — permission denied for schema cutter_governance).
⇒ The exact source-document-version candidate cannot be enumerated from
this macro's read-only path. It needs a privileged read as cutter_exec
(see §5). This is the exact reason a concrete registry candidate is not
named here — not an absence of candidates.
3. Target acceptance criteria (GAP8)
A document is an acceptable first orchestrator-managed production-run target iff ALL hold:
C1 not ICX-CONST — unless an explicit sovereign approval names it
(ICX-CONST is already enacted; re-cutting it is
out of scope for a first run)
C2 has a source snapshot — a pinned normalized snapshot artifact exists
(the cutplan input; sha-addressable)
C3 dry-runnable first — Mode.DRYRUN over the snapshot completes green
BEFORE any Mode.LIVE attempt
C4 small — 1–3 information units for the first run
(bounds blast radius; bounds Compensation A)
C5 rollback feasible — every resulting IU stays 'draft' until a separate
enact decision; Compensation A/B (Report 04) apply
C6 low stakes — not a load-bearing legal instrument; ideally a
synthetic/fixture document authored for the run
C7 fresh change_set — a new change_set_id (not the M1 pinned id) — which
also exercises the F3 generalization (Report 05)
4. Candidate assessment
ICX-CONST: REJECT — C1 (already enacted; sovereign-named only)
test / pilot.*: WEAK — likely C2-fail (ad-hoc rows, provenance unclear); a
privileged read is needed to confirm a source snapshot
D38-DIEU*: UNKNOWN — these are draft IUs already in public.information_unit;
whether each maps to a registered source_document_version
with a snapshot needs the §5 privileged read. If one does
and is 1–3 units, it satisfies C4/C5/C6.
RECOMMENDED: a NEW synthetic low-stakes document authored specifically for the
first run — 1–3 units, deterministic snapshot — best satisfies
C1–C7 and keeps the first run fully controlled.
5. Privileged discovery query (operator / next macro — cutter_exec)
-- run as cutter_exec (creds: DOT_CUTTER_EXEC_* in /opt/incomex/docker/.env)
SELECT v.id AS source_version_id, d.id AS source_document_id,
d.canonical_key, v.snapshot_sha256, v.created_at
FROM cutter_governance.source_document_version_registry v
JOIN cutter_governance.source_document_registry d
ON d.id = v.source_document_id
WHERE d.canonical_key NOT LIKE 'ICX-CONST%'
ORDER BY v.created_at DESC;
-- a row with a non-null snapshot_sha256 and an estimated 1-3 unit cut
-- satisfies C2/C4 and becomes the dry-run candidate.
6. Target creation / intake package (if no registry candidate qualifies)
1. Author a small normalized snapshot artifact (1-3 law units, BEGIN/END
sentinels per the dryrun region contract) — synthetic, low-stakes.
2. Register it: insert source_document_registry + source_document_version_registry
rows (cutter_exec) — this is registration metadata, NOT an IU cut.
3. Mode.DRYRUN the orchestrator over the snapshot → cutplan + write-set preview;
confirm green, capture writer_digest.
4. Pair the dry-run-proven target with the GAP7 approval package (Report 07)
for the first authorised Mode.LIVE run.
Step 2 (registry insert) is content/intake metadata, not an IU production mutation, and is not performed by this macro.
7. Verdict
discovery: DONE — 158 IUs surveyed; ICX-CONST enacted; 98 draft
registry_candidate: NOT NAMED — exact reason: cutter_governance is RO-privilege
denied; needs the §5 cutter_exec query
acceptance_criteria: DEFINED — C1..C7
recommendation: new synthetic low-stakes 1-3-unit document, or a §5
registry hit confirmed by the next macro
intake_package: PRODUCED (§6) — no production mutation performed
g6: PASS