fix(OMN-15959): delete zombie handler_llm_cli_subprocess.py from node_llm_inference_effect

[jonahgabriel]

OMN-15959 — Delete zombie handler_llm_cli_subprocess.py

OMN-13250 (Phase D, Done 2026-07-15) claimed node_llm_inference_effect had been
stripped to HTTP-only and handler_cross_cli_invoker.py deleted. Neither claim
was true. Live state at omnibase_infra@52d8b601c (2026-08-12): the file was
still present (12,022 bytes, mtime 2026-07-26), still exposed
inference.gemini_cli/inference.claude_cli/inference.opencode_cli, still
carried 40 passing tests. OMN-13215 had already killed the architecture this
handler served — omnimarket/src/omnimarket/configs/routing_tiers.yaml:27:
"shelled codex-cli/cli_agents tiers were REMOVED" — the handler was the
stump. Evidence-Source: docs/plans/2026-08-12-delegation-exec-backends-revival-brief.md §1.2 (omni_home).

What changed

• Deleted src/omnibase_infra/nodes/node_llm_inference_effect/handlers/handler_llm_cli_subprocess.py (316 lines) and its 3 tests files (40 tests: test_handler_llm_cli_subprocess.py 4, _claude_opencode.py 8, _execution.py 28).
• Removed the 3 CLI operations (inference.gemini_cli/claude_cli/opencode_cli) from node_llm_inference_effect/contract.yaml handler_routing.handlers; bumped contract_version 1.4.4 → 1.5.0 (operation-surface removal); updated: 2026-08-12.
• Removed the now-dead validation_exemptions.yaml entry keyed to the deleted file.
• Fixed a stale doc-drift finding in .github/workflows/hostile-reviewer.yml (a required CI gate, "Hostile Review Gate"): its header comment and PR-comment template claimed the adversarial review path runs "via HandlerLlmCliSubprocess". Verified against omniintelligence/src/omniintelligence/review_pairing/cli_review.py live: zero references to handler_llm_cli_subprocess/HandlerLlmCliSubprocess anywhere in omniintelligence — the module calls model endpoints directly and only shells out to gh for PR-diff fetching. The claim was never true; corrected in place rather than left dangling post-deletion.
• Updated docstrings in handler_coding_agent_invoke.py, handler_wiring.py (2 spots), and enum_cli_backend_status.py that named the module — either marking it explicitly deleted or genericizing the illustrative example.
• Repointed 2 dead-file references (test_validate_pr_contract_sync.py fixture path, 2 compliance test methods removed from test_handler_autowiring_compliance.py/test_handler_autowiring_compliance_followup.py) and genericized 2 test fixtures in test_wiring.py that used the deleted handler's class name / exact operation strings as illustrative examples for the OMN-9461 "shared handler, multiple operations" ID-collision regression guard — the guarded behavior (unrelated to this handler) still has full coverage under synthetic names.
• Left untouched, deliberately: contracts/OMN-10137.yaml (2 hits) — a per-repo deploy-evidence historical receipt tied to the original OMN-10137 PR that registered these ops; historical record-keeping, not a live reference (same exemption class as OCC ticket-ref retention).

Net: -1396/+53 lines, zero new surface.

Seams

What consumed the 3 deleted operations? Nothing — proven, not assumed.

• omnimarket/src/omnimarket/configs/routing_tiers.yaml:27 — the delegation-tier config that would have selected these ops (cli_agents tier) was already removed under OMN-13215, before this PR. Reconfirmed live on omnimarket@main at PR time.
• Repo-wide positive-controlled grep (post-deletion, this worktree) for the handler symbol (handler_llm_cli_subprocess/HandlerLlmCliSubprocess) and the 3 op names (inference.claude_cli/inference.gemini_cli/inference.opencode_cli) returns zero hits outside: (a) this PR's own "was deleted" annotations, and (b) the deliberately-preserved contracts/OMN-10137.yaml historical receipt.
• pyproject.toml entry points and docker/catalog/ service manifests: zero references to the module or class (checked live, pre-deletion — nothing to unregister).
• RegistryInfraLlmInferenceEffect.get_supported_operations() already only ever declared ["inference.openai_compatible"] — never registered the CLI ops, so no DI-registration code needed touching.

OMN-14355 canon-shape ratchet — verified live, question is moot

Ran omnibase_core/scripts/ci/canonical_handler_shape.py --package omnibase_infra --src-root <this worktree>/src --json (full classification, 116 nodes) against the post-deletion tree:

• Zero occurrences of llm_cli_subprocess anywhere in the classification output.
• node_llm_inference_effect classifies is_canonical: true, handler_module: ...handler_llm_inference_command (unchanged from pre-deletion — classify_node() only ever resolved handler_routing.handlers[0], so it was structurally blind to the CLI handler before deletion too — confirmed by grep on canonical_handler_shape_baseline.py, zero hits for this handler in the committed core baseline).
• The deletion makes the question moot: there is no longer a file for the ratchet to classify or misclassify.

agent_invocation: true count

grep -rn "agent_invocation" src/omnibase_infra/nodes/*/contract.yaml post-deletion: exactly one — node_coding_agent_invoke_effect/contract.yaml:72. The other 3 coding-agent-quartet nodes declare it false; node_remote_agent_invoke_effect has an unrelated operation literally named remote_agent_invocation (not the descriptor boolean).

Test evidence

• tests/unit/nodes/node_llm_inference_effect/ (full node suite): 462 passed, 0 failed.
• tests/unit/runtime/auto_wiring/ + tests/unit/nodes/node_coding_agent/ + tests/scripts/test_validate_pr_contract_sync.py: 554 passed, 0 failed.
• tests/integration/{test_projection_handler_wiring_runtime_dispatch,runtime/test_handler_wiring_resolver_integration,runtime/test_handler_wiring_async_incompat_quarantine_integration,runtime/test_handler_wiring_entry_keys_integration,runtime/test_llm_inference_contract_runtime_bus,projectors/test_handler_wiring_projection_dispatch,handlers/test_handler_autowiring_compliance,test_handler_autowiring_compliance_followup}.py: 28 passed, 1 skipped (Kafka integration test, opt-in by design — KAFKA_INTEGRATION_TESTS=1).
• tests/unit/ full tree (-n auto): 23300 passed, 40 skipped (all pre-existing environment-conditional skips — omnimarket-sibling-not-installed, Docker-daemon-required, flock-unavailable — none touch this change).
• tests/ci/ tests/scripts/ tests/audit/ tests/services/ tests/replay/ tests/chaos/ tests/performance/: 2 pre-existing failures, both orthogonal and unrelated to this diff (verified: zero files in this PR touch either surface):
  • tests/ci/test_env_parity.py::test_k8s_bound_keys_are_bound_in_compose — omnimarket k8s-vs-compose env-key parity drift (KAFKA_CONSUMER_GROUP, PROJECTION_RUNNER_HEALTH_PORT on the projection-writer deployments).
  • tests/performance/test_llm_endpoint_slo.py::TestCoder14BSlo::* — live GPU LLM endpoint returns HTTP 404 for Qwen/Qwen2.5-Coder-14B-Instruct (endpoint-availability test, not code).
• Full tests/integration/ under -n auto showed additional failures under host contention (Unclosed AIOKafkaProducer — Kafka-connectivity signature); none named llm_inference/cli_subprocess/coding_agent in the failing set, and the directly-relevant integration coverage (listed above) is 100% green in isolation — attributed to live-service contention on the local dev box, not this diff. CI's governed selector is the authoritative gate.
• Pre-push governed selector (OMN-13973) escalated to the full unit suite automatically (shared-module touch: handler_wiring.py) and passed: 23300 passed, 40 skipped, ... in 658.52s.
• uv run ruff format --check / uv run ruff check: clean.
• pre-commit run (full hook chain, on commit): all hooks passed.

DoD

• Ticket: OMN-15959 (parent epic OMN-13245).
• proof_class: receipt-bound (grep evidence + CI citations above; gh pr checks will be the final live proof).

Evidence-Source: OCC#6651
Evidence-Ticket: OMN-15959

[coderabbitai]

Warning

Review limit reached

You’ve reached a temporary PR review limit under our Fair Usage Limits Policy.

Your current included review allowance is based on your included PR review attempts over the past 7 days.

Next review available in: 38 minutes

Limit details: You’ve used the included review currently available. Your 125 included PR review attempts over the past 7 days set your current allowance at 1 review per hour.

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.
You're only billed for reviews past your plan's rate limits ($0.25/file).

How can I continue?

Wait for the limit to reset, then comment @coderabbitai review or push new commits to the PR.

An organization admin can change what happens after included review limits in Billing.

How do review limits work?

CodeRabbit enforces per-developer PR review limits within each organization.

For paid Pro and Pro+ reviews, CodeRabbit uses a developer's included PR review attempts over the past 7 days to set the current hourly allowance. At typical activity levels, the full plan allowance applies. Higher sustained activity can lower the allowance until earlier attempts leave the 7-day window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 80995376-13dd-4268-b916-a3fb537d5d1e

📥 Commits

Reviewing files that changed from the base of the PR and between 0ff5b8e and caf6280.

📒 Files selected for processing (20)

• .github/workflows/hostile-reviewer.yml
• docs/evidence/OMN-12548/determinism-audit.json
• scripts/ci/infra-node-allowlist.txt
• src/omnibase_infra/models/coding_agent/enum_cli_backend_status.py
• src/omnibase_infra/nodes/node_coding_agent_invoke_effect/contract.yaml
• src/omnibase_infra/nodes/node_coding_agent_invoke_effect/handlers/handler_coding_agent_invoke.py
• src/omnibase_infra/nodes/node_llm_inference_effect/contract.yaml
• src/omnibase_infra/nodes/node_llm_inference_effect/handlers/handler_llm_cli_subprocess.py
• src/omnibase_infra/runtime/auto_wiring/handler_wiring.py
• src/omnibase_infra/validation/validation_exemptions.yaml
• tests/fixtures/dispatch_parity/baseline-selection-v2.json
• tests/integration/handlers/test_handler_autowiring_compliance.py
• tests/integration/test_handler_autowiring_compliance_followup.py
• tests/integration/test_omn12816_main_promotion_coverage.py
• tests/integration/test_omn_6659_type_suppression_contract_touch.py
• tests/scripts/test_validate_pr_contract_sync.py
• tests/unit/nodes/node_llm_inference_effect/handlers/test_handler_llm_cli_subprocess.py
• tests/unit/nodes/node_llm_inference_effect/handlers/test_handler_llm_cli_subprocess_claude_opencode.py
• tests/unit/nodes/node_llm_inference_effect/handlers/test_handler_llm_cli_subprocess_execution.py
• tests/unit/runtime/auto_wiring/test_wiring.py

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

❓ Hostile Reviewer — UNKNOWN

Blocking findings (critical): 0
Total findings: 0
Models succeeded: none

---

Gate semantics (pilot phase)

Verdict	Meaning	Blocks merge?
passed	No critical findings	No
blocked	CRITICAL findings found	Yes
degraded	All models unavailable (infra)	No (pilot)

Powered by omniintelligence.review_pairing.cli_review — multi-model adversarial review (OMN-8468/OMN-8524)

[onexbot-occ-writer]

OCC autobind did not mint a companion for this PR: no changed-file candidate could be proven RED against the merge base, and emitting a PR-existence probe instead would be non-falsifiable evidence (OMN-15247). Hand-authored evidence is required.