[jonahgabriel]

Evidence-Source: OCC#8985
Evidence-Ticket: OMN-18135

OMN-18135 AC3 + AC4 — the closer's half

Implements the ruling at docs/tracking/ROLLING_WORK_LEDGER.md:6148, 2026-09-10T17:50:03Z. Provenance stated as it is: an orchestrator ruling under the deterministic-truth doctrine, not an operator ruling.

omnimarket#2460 carries the proof class. This carries the join, because deciding whether a criterion is state-shaped needs the ticket's acceptance text, and node_dod_verify has no Linear access and never sees it. That is why AC4 could not be a single omnimarket change.

It moves in both directions

Looser. A ticket whose every parsed criterion asserts live state, each discharged by a verified readback, no longer dies on behavior_proving_count <= 0. This is the case the ruling exists to admit. It is bounded to all criteria: one behaviour-shaped criterion, or a body whose criteria cannot be read at all, and the behaviour proof is still required. A ticket is judged by its strictest criterion, never on average.

Stricter. A behaviour-shaped criterion whose only verified checks are readbacks now holds. Binding used to key on check status alone and never looked at proof class, so a readback silently discharged a criterion asserting what the code does. That hole is invisible to every test that has no behaviour proof at all, because the behaviour conjunct catches those first — so it has its own test with a real behaviour proof sitting beside it, and a positive control where the same shape with a state-shaped criterion does release.

Untouched, deliberately. A criterion proved by a merge-state, surrogate or indeterminate check binds exactly as before. Narrowing those is a different argument nobody has made, and making it here would retroactively un-close tickets that have already flipped on that basis.

AC3: the hold names the bar

The gap_no_behavior_proof comment now says what would clear it:

• which command shapes prove behaviour — the test runners and the ONEX CLI, named individually;
• which prefixes are transparent, so an author does not lose the class to env -u or cd;
• that a live probe through ssh, aws, kubectl, curl, psql or a shell loop is not behaviour-proving, however real the thing it exercised;
• that an asserted readback is the admissible route for a criterion asserting live state, and that the command must be able to go red on what it read.

Three lanes spent a night rediscovering this rule by measurement because the comment stated a verdict and no remedy.

The honest limit, in the code and not only here

Criterion shape is read from prose. It is a heuristic and it is tight on purpose: state-shaped requires a positive state marker and no behaviour marker, and everything else falls to behaviour and holds. The asymmetry is chosen rather than incidental — reading a criterion as behaviour-shaped costs a comment, while reading one as state-shaped releases a flip on a readback. So the vocabulary earns its way in one phrase at a time, every entry is drawn from a criterion that exists in the corpus, and the veto wins ties. test_anything_else_falls_to_behaviour_and_holds pins the ambiguous cases, including a state marker vetoed by behaviour language.

RED first

At parent 0a704957c, verbatim: 13 failed, 4 passed. After: 19 passed in the new module and 416 passed across the closer suite.

Verification

• Governed pre-push green on 9fddfcc8b8e82b0ba524dfeb7d3125d72d895041. The selector escalated and placed the run off-box: REMOTE LAB RUN PASS accepted ... h201.3 ran 27597 tests green ... (suite log sha256 6009ab599f801d9e003e31d15728e586fe718b2cbd590d0409772d33f68c6151, 285s). No override variable set, no escalation bypassed.
• uv run mypy --strict on the node: clean. ruff format --check, ruff check: clean.
• node_version 1.11.0 to 1.12.0, so the gap-comment fingerprint moves with the rule it describes.
• Lab coverage, honestly: this node runs only inside the scheduled evidence-autoclose-sweep workflow, not on the .201 compose dev lane or the onex-lab overlay. No lab pass exercises it and none is claimed. Its proof surface is the next scheduled closer tick after merge.

Landing order

omnimarket#2460 should land first. It is inert on its own — binding already keyed on status and never on proof class, and behavior_proving_count does not change — so nothing moves until this PR teaches the closer the join. Landing this one first would be the harmless order too, since readback_proving_count reads 0 from a verifier that does not emit it and 0 releases nothing.

[github-actions]

Hostile Reviewer — adversarial findings (OMN-17492)

Models succeeded: glm-review
Models failed: codex
New finding threads: 1
Deduped (already posted on this PR): 0
Nit-level findings suppressed: 1

The model is the FINDER, never the gate: merge is gated only by the
deterministic Hostile Review Thread Gate, which blocks while
hostile-reviewer threads are unresolved. Resolve each thread after
addressing (or rejecting, with a reply) its finding.

Findings not anchored to a changed file

• [MAJOR] hostile-reviewer (glm-review)

  Overly broad state markers ('returns', 'responds', 'serves', 'contains') can release behaviour criteria | _STATE_MARKER_RE admits \breturns?\b, \bresponds?\b, \bserves?\b, and \bcontains?\b as positive evidence of live-state shape. These verbs are equally at home in behaviour assertions, e.g. 'the API returns 404 when the token is expired', 'the parser serves the malformed case', 'the CLI contains a retry loop'. None of these examples carry a veto word from _BEHAVIOUR_MARKER_RE (no 'test', 'handler', 'suite

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

• [MAJOR] hostile-reviewer (glm-review)

  State-shaped gate trusts parsed criteria only; criteria outside the parser are unchecked | _every_criterion_is_state_shaped gates the behavior_proving_count conjunct on _acceptance_criteria_items(description). If the parser extracts a subset of the ticket's real acceptance criteria (e.g. criteria listed without a recognised heading, nested lists, or a body over a truncation limit), the join is computed over a partial view. A ticket with one state-shaped parseable criterion and one unparseable behaviour crit

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

• [MINOR] hostile-reviewer (glm-review)

  Readback-blocked note truncates without an overflow indicator | readback_note joins readback_blocked[:_MAX_UNCOVERED_LISTED] with no 'and N more' suffix. The unbound list elsewhere in the same function tracks unlabelled counts and appends summary text; the new note can silently under-report when more criteria than _MAX_UNCOVERED_LISTED are readback-blocked, which understates the scope of the hold the author is being asked to fix. | Evidence: " OMN-18135: "
  + ", ".join(readback_blocked[:_MAX_UNCO

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

• [MINOR] hostile-reviewer (glm-review)

  Admissible-command list duplicated as prose in the hold comment will drift | The gap_no_behavior_proof reason enumerates test runners, transparent prefixes, and readback-capable tools as a hardcoded English sentence. The classifier that actually decides proof class lives elsewhere (omnimarket / the verifier). When the verifier adds a runner or prefix, the comment becomes false, and nothing in this repo pins the two together. The diff bumps _GAP_FINGERPRINT_CONTRACT_VERSION for the gap fingerprint but no tes

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

Findings demoted from threads (anchor rejected)

• [MAJOR] hostile-reviewer (glm-review)

  No test for readback release with a description lacking any criteria | The comment on the new conjunct claims a body with no parseable criteria is held by the OMN-18056 binding gate 'a conjunct later'. This dependency is asserted in prose but untested in the diff. If _ac_binding_gap permits a description with no AC heading through some other leg (for example when bindings exist for labels a human added to the companion contract rather than the ticket body), a readback_proving_count > 0 combined with an empt

  Resolve this thread when addressed — the Hostile Review Thread Gate blocks while hostile-reviewer threads are unresolved (OMN-17492).

[github-actions]

⚠️ Hostile Reviewer — DEGRADED (informational)

Critical findings: 0
Major findings: 3
Total findings: 6
Models succeeded: glm-review

Note: Fewer than 2 reviewer models succeeded. Degraded results are informational (OMN-8468/OMN-8524) and do not block merge. Error: cli_review exit 2 (fewer than 2 models succeeded — partial/total outage)

---

Semantics (OMN-17492 — the model finds, thread resolution gates)

Surface	Meaning	Blocks merge?
Review threads	Per-finding, posted by the reviewer	No (informational)
Hostile Review Thread Gate	Deterministic: unresolved hostile-reviewer threads exist	Fails until resolved (not yet a required context)
degraded verdict	Fewer than 2 models succeeded (infra)	No

Powered by omniintelligence.review_pairing.cli_review — multi-model adversarial review: qwen3-review, qwen3-review-b, glm-review (OMN-8468/OMN-8524/OMN-17492)