[jonahgabriel]

A delegation that fully succeeds exits non-zero. This makes the exit status agree with the terminal.

What happens today

_require_completed_terminal_evidence documents itself as refusing "a completed receipt that lacks evidence its request required", and its only caller passed both flags as literal True. So every completed delegation was required to carry budget evidence and response-contract evidence, which no caller asked for and the deployed producer emits for neither.

Measured on the dev lane, correlation 0b21b5f0-fb56-4ab6-9219-24bbfd841f35: the terminal was completed, the quality gate passed at 1.0, the projection row was written, and the CLI still exited non-zero. Nothing about the delegation failed; the receipt assertion did.

That is not cosmetic here. Every lane brief in this session cites a delegate receipt, so a caller branching on exit status reads success as failure, while a caller reading the terminal reads it correctly. Two readers of the same run disagreeing is the defect.

What changes, and what deliberately does not

The check itself is untouched and still refuses. What changes is who arms it. A response contract in the request arms the contract half; nothing arms it when no contract was asked for, because a run that asked for no contract has no contract to evidence and refusing it asserts a requirement nobody stated.

The budget half returns false, and the reason is worth stating rather than leaving as a default. Nothing a caller can pass expresses "I require budget evidence": the comparison happens only where the backend declares its own input-token budget, which is not knowable when the request is built. There is no honest request-side predicate to return. It is returned as a value rather than dropped, so the day a demand exists it has exactly one place to land, and the refusal stays armed either way.

Why this cannot have quietly become a no-op

That is the obvious risk in relaxing a gate, so it is the thing under the most test. Three positive controls assert the refusal still fires: on the contract half alone, on the budget half alone, and on both together with each named in one message. A fourth pins that a non-completed terminal is still never refused here, so the guard cannot widen onto it.

The derivation is also read directly rather than only through behaviour, so an edit back to a literal True is a red test rather than another silent outage.

Context

Third surface of the same change as the two fixed under the parent, all three from #3882: a consumer-side requirement landing ahead of the producer that satisfies it. The first refused the CLI before dispatch, the second terminalized every lane delegation, and this one only disagrees about the exit code. Severity descends across the three and they should not be read as equivalent.

Local: 746 tests green in tests/unit/cli/, mypy --strict clean, ruff clean, whole-tree hooks green, runner image identity regenerated because the version bump restales it.

Owed

The lane readback: a fresh delegate whose exit status is zero on a completed terminal. It needs this merged and the dispatch venv reconciled, and it is cited on the ticket when taken.

Evidence-Ticket: OMN-18956
Evidence-Source: OCC#10616

[github-actions]

Hostile Reviewer — adversarial findings (OMN-17492)

Models succeeded: glm-review
Models failed: codex
New finding threads: 0
Deduped (already posted on this PR): 0
Below quorum (one model only, reported not threaded): 5
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.

Below quorum: 5 finding(s) raised by one model only (OMN-18479)

These are reported and NOT dropped, but they get no thread and do not block: a single model's finding no other model reproduced is not evidence enough to stop a merge. Read them; act on them if they are right.

• [CRITICAL] src/omnibase_infra/cli/cli_delegate.py, _receipt_evidence_requirements (glm-review) — Contract evidence demanded but producer reportedly never emits it | The docstring states the deployed producer emits neither budget_evidence nor response_contract_evidence. Yet _receipt_evidence_requi
• [MAJOR] tests/unit/cli/test_receipt_evidence_only_when_requested_omn18956.py (glm-review) — No test exercises the new call-site wiring | The tests validate _require_completed_terminal_evidence (unchanged code) and _receipt_evidence_requirements in isolation. Nothing verifies that run_delegat
• [MINOR] src/omnibase_infra/cli/cli_delegate.py (glm-review) — Dead parameter retained on speculative future demand | require_budget_evidence is threaded through _delegate_receipt_evidence_error and the call site while the derivation always returns False. The doc
• [MINOR] src/omnibase_infra/cli/cli_delegate.py, _receipt_evidence_requirements (glm-review) — Any non-None response_contract arms the check, including empty or placeholder contracts | The predicate treats dict as a boolean proxy for 'a contract was asked for.' A caller passing an empty dict {}
• [MINOR] docker/runners/runner-image.lock.json (glm-review) — Lock digests rotated in same PR with no stated provenance | identity_digest and shared_env_digest change with no accompanying change to the inputs (shared_env_install_args is unchanged) and no comment

[github-actions]

⚠️ Hostile Reviewer — DEGRADED (informational)

Critical findings: 0
Major findings: 3
Total findings: 7
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)

[github-actions]

Hostile Reviewer — adversarial findings (OMN-17492)

Models succeeded: glm-review
Models failed: codex
New finding threads: 0
Deduped (already posted on this PR): 0
Below quorum (one model only, reported not threaded): 5
Nit-level findings suppressed: 2

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.

Below quorum: 5 finding(s) raised by one model only (OMN-18479)

These are reported and NOT dropped, but they get no thread and do not block: a single model's finding no other model reproduced is not evidence enough to stop a merge. Read them; act on them if they are right.

• [MAJOR] src/omnibase_infra/cli/cli_delegate.py, _receipt_evidence_requirements (glm-review) — Budget evidence enforcement is now globally unreachable | The previous behavior enforced budget_evidence on every completed terminal. The new derivation returns require_budget_evidence=False unconditi
• [MAJOR] tests/fixtures/delegation/omn18956/completed_terminal_no_evidence.json (glm-review) — Frozen capture fixture embeds infrastructure details and credential material | The verbatim log capture committed as a test fixture contains the tailnet hostname and Kafka broker endpoint (omninode-pc
• [MAJOR] src/omnibase_infra/cli/cli_delegate.py (glm-review) — The production wiring path is never tested | The integration tests were written precisely because the failure occurred 'one layer out', at the receipt-validator wiring. Yet no test exercises run_deleg
• [MINOR] src/omnibase_infra/cli/cli_delegate.py, run_delegate (glm-review) — Tuple index access for a two-value semantic pair | receipt_evidence_demanded[0] and [1] at the call site rely on positional knowledge of an undocumented tuple. A future reorder or third element breaks
• [MINOR] src/omnibase_infra/cli/cli_delegate.py, _receipt_evidence_requirements (glm-review) — Empty dict response_contract arms the contract-evidence requirement | The predicate is response_contract is not None, so response_contract={} arms the requirement. If an empty dict is a legitimate 'no