omnilane 0.42.1 → 0.42.2

package/.claude-plugin/marketplace.json

@@ -6,14 +6,14 @@
   },
   "metadata": {
     "description": "One routing table, native-first or CLI delegation, supervised completion workflows.",
-    "version": "0.42.1"
+    "version": "0.42.2"
   },
   "plugins": [
     {
       "name": "omnilane",
       "source": "./",
       "description": "Route each subtask to the model that fits, then delegate through a compatible caller-owned native agent or vendor CLI.",
-      "version": "0.42.1",
+      "version": "0.42.2",
       "category": "development",
       "keywords": [
         "routing",

package/.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "omnilane",
-  "version": "0.42.1",
+  "version": "0.42.2",
   "description": "One routing table, every harness: classify subtasks into lanes and delegate through compatible caller-owned native agents or vendor CLIs with exact-AA downward policy and supervised jobs.",
   "author": {
     "name": "Seraphim0916"

package/CHANGELOG.md

@@ -6,6 +6,18 @@ semantic version tags.
 
 ## [Unreleased]
 
+## [0.42.2] - 2026-09-07
+
+### Fixed
+
+- Pass explicit Grok effort through `--reasoning-effort` instead of silently discarding it. Accept the CLI's `low`, `medium`, `high`, and `xhigh` selectors; reject invalid values before starting the provider.
+- Require a verified `cli_reasoning_effort` transport contract with the exact CLI flag before admitting scored Grok targets. Host-local overlays retain their host, snapshot, exact identity, and evidence-hash checks; frozen AA scores and the approved registry SHA remain unchanged.
+- Make Grok 4.6 routing defaults explicitly `high`. Missing or stale mappings still fail closed, and explicit effort is rejected on the unsupported live ACP path.
+
+### Documentation
+
+- Document the distinction between request-selector verification and upstream model identity, and the local overlay refresh required when the CLI or runner changes.
+
 ## [0.42.1] - 2026-09-07
 
 ### Fixed
@@ -837,7 +849,8 @@ work to the wrong model, and records the evidence behind the shipped defaults.
 - Initial shared routing table, cross-vendor dispatcher, runners, installer,
   and baseline lint fixes.
 
-[Unreleased]: https://github.com/Seraphim0916/omnilane/compare/v0.42.1...HEAD
+[Unreleased]: https://github.com/Seraphim0916/omnilane/compare/v0.42.2...HEAD
+[0.42.2]: https://github.com/Seraphim0916/omnilane/compare/v0.42.1...v0.42.2
 [0.42.1]: https://github.com/Seraphim0916/omnilane/compare/v0.42.0...v0.42.1
 [0.42.0]: https://github.com/Seraphim0916/omnilane/compare/v0.41.1...v0.42.0
 [0.41.1]: https://github.com/Seraphim0916/omnilane/compare/v0.40.0...v0.41.1

package/README.ja.md

@@ -533,6 +533,12 @@ work の別名ではありません。サービス管理など、work の境界
 
 ## 📜 リリース履歴
 
+## v0.42.2 の新機能
+
+- **Grok の推論強度を CLI に渡します。** 明示的な `low`、`medium`、`high`、`xhigh` は `--reasoning-effort` で渡され、Grok 4.6 の既定ルートは `high` を選択します。
+- **証拠に基づくローカル対応付け。** ホストローカルの overlay で正確な CLI セレクター契約を検証し、固定 AA スコアや承認済み registry SHA は変更しません。対応付けの欠落・誤りは引き続き拒否し、live ACP の明示的な強度指定も検証完了まで拒否します。
+- **更新。** npm 公開後に `npm i -g omnilane@0.42.2` を実行します。既存の repo-symlink インストールは checkout を更新し、再インストールせずに `omnilane --version` を確認できます。
+
 ## v0.42.1 の新機能
 
 - **CI フィクスチャの修復。** 完全な Python discovery では、旧 routing／Grok readiness テストに synthetic-human caller を明示します。production の missing-identity 拒否、承認済み registry SHA、下方向スコア検査、retry lineage、skip assertion は変更しません。

package/README.ko.md

@@ -518,6 +518,12 @@ work 는 지정한 디렉터리 안의 변경만 허용하며 모델 연결은
 
 ## 📜 릴리스 기록
 
+## v0.42.2 새 기능
+
+- **Grok 추론 강도를 CLI에 전달합니다.** 명시적 `low`, `medium`, `high`, `xhigh` 선택을 `--reasoning-effort`로 전달하며 Grok 4.6 기본 경로는 `high`를 선택합니다.
+- **증거 기반 로컬 매핑.** 호스트 로컬 overlay로 정확한 CLI 선택자 계약을 검증하며 고정 AA 점수와 승인된 registry SHA는 변경하지 않습니다. 누락되거나 잘못된 매핑은 계속 거부하며 live ACP의 명시적 강도도 검증될 때까지 차단합니다.
+- **업그레이드.** npm 게시 후 `npm i -g omnilane@0.42.2`를 실행합니다. 기존 repo-symlink 설치는 checkout을 업데이트하고 재설치 없이 `omnilane --version`을 확인할 수 있습니다.
+
 ## v0.42.1 새 기능
 
 - **CI 픽스처 복구.** 전체 Python discovery는 기존 routing 및 Grok readiness 테스트에 synthetic-human caller를 명시합니다. production의 missing-identity 거부, 승인된 registry SHA, 하향 점수 검사, retry lineage, skip assertion은 변경하지 않습니다.

package/README.md

@@ -627,6 +627,12 @@ working notes, including per-benchmark caveats, live in
 
 ## 📜 Release history
 
+## What's new in v0.42.2
+
+- **Grok effort reaches the CLI.** Explicit `low`, `medium`, `high`, and `xhigh` selections are passed with `--reasoning-effort`; Grok 4.6 default routes now select `high`.
+- **Evidence-backed local mapping.** A host-local overlay proves the exact CLI selector contract without changing frozen AA scores or the approved registry SHA. Missing or incorrect mappings still deny dispatch; explicit effort on live ACP remains blocked until that surface is verified.
+- **Upgrade.** After npm publication, run `npm i -g omnilane@0.42.2`. Existing repo-symlink installations can update their checkout and verify `omnilane --version` without rerunning installation.
+
 ## What's new in v0.42.1
 
 - **CI fixture repair.** Full Python discovery now gives legacy routing and Grok-readiness fixtures an explicit synthetic-human caller, while production missing-identity denial, the approved registry SHA, downward score checks, retry lineage, and skip assertions remain unchanged.

package/README.zh-CN.md

@@ -493,6 +493,12 @@ scripts/dispatch.sh --dry-run hardest-coding "…"   # 完整解析后的计划,
 
 ## 📜 版本历程
 
+## v0.42.2 新功能
+
+- **Grok 强度确实传入 CLI。** 显式的 `low`、`medium`、`high`、`xhigh` 通过 `--reasoning-effort` 传递；Grok 4.6 默认路由指定 `high`。
+- **本机映射以证据验证。** 主机本地覆盖文件证明精确 CLI 选择器契约，不调整固定 AA 分数或获准的注册表 SHA。缺失或错误映射仍拒绝派工；实时 ACP 的显式强度在验证完成前仍阻止执行。
+- **升级。** npm 发布后执行 `npm i -g omnilane@0.42.2`。现有 repo-symlink 安装更新 checkout 并确认 `omnilane --version`，无需重新安装。
+
 ## v0.42.1 新功能
 
 - **修复 CI 测试夹具。** 完整 Python discovery 现在会为旧 routing 与 Grok readiness 测试显式指定 synthetic-human caller；production 的缺失身份拒绝、获准 registry SHA、向下分数闸、重试 lineage 与 skip 断言均保持不变。

package/README.zh-TW.md

@@ -534,6 +534,12 @@ scripts/dispatch.sh --dry-run hardest-coding "…"   # 完整解析後的計畫,
 
 ## 📜 版本歷程
 
+## v0.42.2 新功能
+
+- **Grok 強度確實傳入 CLI。** 明示的 `low`、`medium`、`high`、`xhigh` 透過 `--reasoning-effort` 傳遞；Grok 4.6 預設路由指定 `high`。
+- **本機映射以證據驗證。** 主機本機覆寫檔證明精確 CLI 選擇器契約，不調整固定 AA 分數或核准的登錄表 SHA。缺少或錯誤映射仍拒絕派工；即時 ACP 的明示強度在驗證完成前仍封鎖。
+- **升級。** npm 發布後執行 `npm i -g omnilane@0.42.2`。既有 repo-symlink 安裝更新 checkout 並確認 `omnilane --version`，不需重跑安裝。
+
 ## v0.42.1 新功能
 
 - **修復 CI 測試資料。** 完整 Python discovery 現在會讓舊 routing 與 Grok readiness 測試明示 synthetic-human caller；production 的缺少身分拒絕、核准 registry SHA、向下分數閘、重試 lineage 與 skip 斷言都維持不變。

package/VERSION

@@ -1 +1 @@
-0.42.1
+0.42.2

package/docs/release-notes-0.42.2.md

@@ -0,0 +1,31 @@
+# Omnilane 0.42.2
+
+This patch repairs the Grok single-shot reasoning-effort transport. It does not raise caller scores or remove the exact-AA downward-delegation gate.
+
+## Changes
+
+- Explicit Grok effort is forwarded as `--reasoning-effort VALUE`. The supported selector spellings are `low`, `medium`, `high`, and `xhigh`; invalid values fail before provider startup.
+- Grok 4.6 entries in the default routing table explicitly select `high`.
+- A scored Grok target requires a verified `cli_reasoning_effort` mapping and the exact `--reasoning-effort` flag. The existing host-local transport overlay validates host, snapshot, exact identity, evidence hashes, vendor, flag, and effort.
+- Explicit effort on the live ACP path is rejected rather than discarded. This release does not add Grok work-mode network isolation on macOS.
+
+## Verification boundary
+
+The frozen AA registry and its approved SHA remain unchanged. Transport evidence belongs to the local host; the package does not ship a blanket assertion that every Grok model/effort combination is verified.
+
+`request-selector-contract` proves how Omnilane selects the model and effort. It does not independently authenticate the upstream model's internal identity: `upstream_identity_verified` remains false. A successful reply alone is not an identity attestation.
+
+For an existing local overlay, use `selector_type: cli_reasoning_effort`, `cli_flag: --reasoning-effort`, and matching model/effort identity only after checking the installed CLI and runner. Retain absolute evidence paths and SHA-256 hashes. Changed evidence requires re-verification; do not merely relabel an old mapping as verified. Select it with `--transport-overlay /absolute/overlay.json` or `OMNILANE_AA_TRANSPORT_OVERLAY`.
+
+## Upgrade
+
+After npm publication:
+
+```sh
+npm i -g omnilane@0.42.2
+omnilane --version
+```
+
+Repo-symlink installations use the updated checkout. Do not rerun `install.sh` just to update the version. Local routing overrides take precedence; review any Grok entries still using `-`.
+
+GitHub release, npm publication, local provider smoke, and Linux CI are separate verification surfaces. Release evidence must identify each result rather than equating one with the others.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnilane",
-  "version": "0.42.1",
+  "version": "0.42.2",
   "description": "One routing table, every harness — classify subtasks into lanes and delegate each lane through a compatible caller-owned native agent or vendor CLI.",
   "bin": {
     "omnilane": "bin/omnilane"
@@ -29,7 +29,7 @@
     "docs/model-capabilities-2026-09.md",
     "docs/native-executor.md",
     "docs/completion-wakeup.md",
-    "docs/release-notes-0.42.1.md",
+    "docs/release-notes-0.42.2.md",
     "hooks/",
     "skills/",
     ".claude-plugin/",

package/plugin.json

@@ -1,6 +1,6 @@
 {
   "$schema": "https://antigravity.google/schemas/v1/plugin.json",
   "name": "omnilane",
-  "version": "0.42.1",
+  "version": "0.42.2",
   "description": "One routing table, every harness: classify subtasks into lanes and delegate through compatible caller-owned native agents or vendor CLIs with exact-AA downward policy and supervised jobs."
 }

package/routing.yaml

@@ -3,7 +3,7 @@
 #   vendor  = codex | claude | grok | gemini | kimi | qwen | opencode | openrouter
 #   openrouter is direct-API (needs only OPENROUTER_API_KEY + curl) and advise/consult
 #   only — it cannot edit files, so it is not in the default work-capable chains.
-#   effort  = codex reasoning effort / claude --effort / agy thinking suffix / (grok: ignored)
+#   effort  = codex reasoning effort / claude --effort / agy thinking suffix / grok --reasoning-effort
 #   Fallback chain: the first candidate whose vendor CLI is installed wins, so the
 #   same table degrades gracefully when you only subscribe to one or two vendors.
 # Override any line in ~/.omnilane/routing.local.yaml (same format; local wins).
@@ -20,17 +20,17 @@
 # standard short-context API tier — on subscription CLIs treat $ as relative ranking.
 # Your own job outcomes (~/.omnilane/jobs/) outrank these priors; edit lanes to match.
 
-hardest-coding: claude claude-fable-5-1 max | codex gpt-6-astra xhigh | grok grok-4.6 - | gemini gemini-3.8-flash-high - # correctness-first coding: Fable max leads the same-condition comparison; Astra xhigh is the default Codex quality/cost tradeoff; explicit --effort max remains available; Grok and Flash keep cross-vendor depth
+hardest-coding: claude claude-fable-5-1 max | codex gpt-6-astra xhigh | grok grok-4.6 high | gemini gemini-3.8-flash-high - # correctness-first coding: Fable max leads the same-condition comparison; Astra xhigh is the default Codex quality/cost tradeoff; explicit --effort max remains available; Grok and Flash keep cross-vendor depth
 bulk-mechanical: codex gpt-5.6-sol high | gemini gemini-3.8-flash-high - | claude claude-sonnet-5 high # endurance and migration work stays on proven Sol high; Flash 3.8 refreshes the fast middle fallback; Sonnet preserves cross-vendor depth
 triage: codex gpt-5.6-luna high | gemini gemini-3.8-flash-low - | claude claude-haiku-4-5 - # first-pass filtering favors the low-cost Luna row; Flash low and Haiku remain cheap cross-vendor fallbacks
-hard-judgment: claude claude-fable-5-1 xhigh | codex gpt-6-astra xhigh | grok grok-4.6 - # strongest same-condition judgment row first, then an independently verified Codex family and Grok; this lane is not a controller selector
-taste-final: claude claude-fable-5-1 xhigh | codex gpt-6-astra xhigh | grok grok-4.6 - | gemini gemini-3.8-flash-high - # final prose and style still require human taste review; general quality evidence orders the candidates without claiming an aesthetic benchmark
-consult: codex gpt-6-astra xhigh | claude claude-fable-5-1 xhigh | grok grok-4.6 - | gemini gemini-3.8-flash-medium - # direct named-model consultation chain; keep --vendor to pin the requested family and prevent fallback
+hard-judgment: claude claude-fable-5-1 xhigh | codex gpt-6-astra xhigh | grok grok-4.6 high # strongest same-condition judgment row first, then an independently verified Codex family and Grok; this lane is not a controller selector
+taste-final: claude claude-fable-5-1 xhigh | codex gpt-6-astra xhigh | grok grok-4.6 high | gemini gemini-3.8-flash-high - # final prose and style still require human taste review; general quality evidence orders the candidates without claiming an aesthetic benchmark
+consult: codex gpt-6-astra xhigh | claude claude-fable-5-1 xhigh | grok grok-4.6 high | gemini gemini-3.8-flash-medium - # direct named-model consultation chain; keep --vendor to pin the requested family and prevent fallback
 ui-draft: codex gpt-5.6-sol high | claude claude-fable-5-1 xhigh | gemini gemini-3.8-flash-high - # UI drafts require a design system or references; measured coding/agent evidence supports the order but does not prove visual taste
 long-context: gemini gemini-3.8-flash-medium - | codex gpt-5.6-terra max | claude claude-opus-5 medium # Flash medium leads long-document synthesis; Terra and Opus stay because context capacity alone does not prove task quality
 fast-agentic: gemini gemini-3.8-flash-low - | codex gpt-5.6-luna high | claude claude-haiku-4-5 - # low-latency tool loops favor Flash low; Luna and Haiku retain cross-vendor fallback depth
-live-search: grok grok-4.6 - | gemini gemini-3.8-flash-high - | claude claude-sonnet-5 high | off # Grok remains the native X/web choice; Flash and Sonnet provide generic web-search fallback, not equivalent social context
-coding-overflow: grok grok-4.6 - | gemini gemini-3.8-flash-high - | kimi kimi-k3 - | qwen qwen3-coder-plus - | opencode - - | off # explicit quota-relief lane; keep all existing non-Codex fallbacks and do not infer unverified Qwen aliases from another harness
+live-search: grok grok-4.6 high | gemini gemini-3.8-flash-high - | claude claude-sonnet-5 high | off # Grok remains the native X/web choice; Flash and Sonnet provide generic web-search fallback, not equivalent social context
+coding-overflow: grok grok-4.6 high | gemini gemini-3.8-flash-high - | kimi kimi-k3 - | qwen qwen3-coder-plus - | opencode - - | off # explicit quota-relief lane; keep all existing non-Codex fallbacks and do not infer unverified Qwen aliases from another harness
 arbitrate: off - - # opinion panel remains opt-in because each voter and round consumes quota
                                                 # Enable: `arbitrate: vote codex,claude,grok -` (any 1-4 of codex/claude/grok/gemini)
                                                 # Debate round (each voter rebuts the others): set the effort field to 2.

package/scripts/lib/aa_policy.py

@@ -164,7 +164,11 @@ def apply_transport_overlay(registry: dict[str, Any]) -> None:
         _check(mapping.get("verification") == "request-selector-contract", "unsupported overlay verification")
         _check(mapping.get("runtime_effort") == row["effort"], "overlay effort mismatch")
         selector_type = mapping.get("selector_type", "model_and_effort")
-        _check(selector_type in ("model_and_effort", "model_id_encoded_effort"), "unknown selector type")
+        _check(selector_type in ("model_and_effort", "model_id_encoded_effort", "cli_reasoning_effort"), "unknown selector type")
+        if selector_type == "cli_reasoning_effort":
+            _check(row["vendor"] == "grok", "unsupported CLI-effort vendor")
+            _check(mapping.get("cli_flag") == "--reasoning-effort", "unproven CLI-effort flag")
+            _check(mapping.get("runtime_effort") in ("low", "medium", "high", "xhigh"), "unsupported CLI effort")
         if selector_type == "model_id_encoded_effort":
             _check(row["vendor"] == "gemini", "unsupported encoded-effort vendor")
             _check(mapping.get("runtime_model") in row["transport_mapping"].get("candidate_model_ids", []), "unproven encoded model selector")
@@ -175,6 +179,7 @@ def apply_transport_overlay(registry: dict[str, Any]) -> None:
             status="verified", runtime_verified=True,
             runtime_model=mapping["runtime_model"], runtime_effort=mapping["runtime_effort"],
             selector_type=selector_type,
+            cli_flag=mapping.get("cli_flag") if selector_type == "cli_reasoning_effort" else None,
             verification="request-selector-contract", upstream_identity_verified=False,
             overlay_sha256=digest, overlay_host=overlay["host"],
         )
@@ -248,10 +253,14 @@ def _runtime_target(registry: dict[str, Any], vendor: str, model: str,
             candidates.append(row)
         else:
             unresolved.append(row["id"])
-    if vendor == "grok" and candidates:
-        # The checked-in Grok runner accepts EFFORT for interface parity but
-        # discards it.  A scored reasoning/effort row therefore cannot be
-        # proven by that runtime surface.
+    if vendor == "grok" and candidates and any(
+        row["transport_mapping"].get("selector_type") != "cli_reasoning_effort"
+        or row["transport_mapping"].get("cli_flag") != "--reasoning-effort"
+        or effort not in ("low", "medium", "high", "xhigh")
+        for row in candidates
+    ):
+        # Only the verified single-shot CLI selector proves scored effort.
+        # Legacy/encoded selectors and live ACP remain unsupported.
         return None, "runtime-effort-discarded", {"vendor": vendor, "model": model, "effort": effort}
     if len(candidates) == 1:
         return candidates[0], "runtime-mapping-verified", {}

package/scripts/runners/run-grok.sh

@@ -2,12 +2,17 @@
 set -euo pipefail
 # omnilane runner: Grok Build CLI
 # Usage: run-grok.sh MODE WORKDIR MODEL EFFORT PROMPT_FILE OUTPUT_FILE
-# EFFORT is accepted for interface parity; Grok has no reasoning-effort knob.
+# Explicit EFFORT is passed through Grok CLI's reasoning-effort selector.
 
 source "$(dirname "${BASH_SOURCE[0]}")/../lib/common.sh"
 
 MODE="$1"; WORKDIR="$2"; MODEL="$3"; EFFORT="$4"; PROMPT_FILE="$5"; OUTPUT_FILE="$6"
-: "$EFFORT" # parity with the uniform runner interface; Grok has no effort knob
+EFFORT_ARGS=()
+case "$EFFORT" in
+  low|medium|high|xhigh) EFFORT_ARGS=(--reasoning-effort "$EFFORT") ;;
+  -|"") ;; # Unspecified effort has no scored runtime mapping.
+  *) echo "omnilane: invalid Grok reasoning effort '$EFFORT'" >&2; exit 2 ;;
+esac
 
 GROK_BIN="${GROK_BIN:-grok}"
 RUN_TIMEOUT="${OMNILANE_TIMEOUT:-600}"
@@ -78,6 +83,10 @@ if [[ "$MODE" != "sysops" && -n "$LIVE_INBOX" ]]; then
   exit 2
 fi
 if [[ -n "$LIVE_INBOX" && -p "$LIVE_INBOX" ]]; then
+  if [[ ${#EFFORT_ARGS[@]} -gt 0 ]]; then
+    echo "omnilane: explicit Grok reasoning effort is not verified for live ACP; use single-shot" >&2
+    exit 2
+  fi
   EVENTS_FILE="${OUTPUT_FILE}.events.jsonl"
   STDERR_FILE="${OUTPUT_FILE}.stderr.log"
   PROGRESS_FILE="${OUTPUT_FILE}.progress.log"
@@ -145,6 +154,7 @@ fi
 ARGS=(--cwd "$WORKDIR" --model "$MODEL"
       --no-memory --no-subagents --no-plan --no-alt-screen
       --output-format plain --verbatim --prompt-file "$PROMPT_FILE")
+[[ ${#EFFORT_ARGS[@]} -eq 0 ]] || ARGS+=("${EFFORT_ARGS[@]}")
 [[ ${#THREAD_ARGS[@]} -eq 0 ]] || ARGS+=("${THREAD_ARGS[@]}")
 ARGS+=("${MODE_ARGS[@]}")
 # Web/X search stays ON by default for advise and sysops.

package/docs/release-notes-0.42.1.md

@@ -1,32 +0,0 @@
-# Omnilane 0.42.1
-
-Omnilane 0.42.1 is a test-fixture and packaging patch for the already-published 0.42.0 exact-AA release. It does not rewrite the `v0.42.0` tag or weaken production routing policy.
-
-## Fixes
-
-- Legacy routing and Grok-readiness tests now explicitly declare a synthetic-human caller. The fixtures no longer depend on ambient caller metadata, while real model calls without exact identity continue to fail closed.
-- The cross-vendor encoded-effort lineage spy uses portable `#!/usr/bin/env python3` and explicit `--background --single-shot` followed by bounded job completion waiting. This isolates the one-shot provider fixture from Gemini's default live/FIFO lifecycle while retaining the exact `--model gemini-3.8-flash-high` selector, model caller/child ceiling, and no-human-exemption assertions.
-- The npm package points at these 0.42.1 notes and retains all five README translations, the AA policy, and the native/completion-wakeup protocol documents.
-
-## Policy boundary
-
-The approved exact-AA registry SHA pin, missing-identity denial, downward score ceiling, child caller context, retry-lineage intersection, and model-retry human-exemption rules are unchanged. Registry accounting remains:
-
-- 78 scored eligible configurations;
-- 1 scored reference-only comparison entry;
-- 10 unknown configurations.
-
-## Upgrade
-
-After npm publication:
-
-```sh
-npm i -g omnilane@0.42.1
-omnilane --version
-```
-
-For an existing repo-symlink installation, update the checkout and run `omnilane --version`. Do not rerun `./install.sh` unless intentionally reviewing and changing integration wiring. A GitHub release does not by itself prove npm publication.
-
-## Verification target
-
-The patch release gate is the complete CI Python discovery command, the full shell suite, package/release policy checks, and a smoke test of the CLI extracted from the built npm tarball. Local preparation records are not packaged release evidence. Published-platform verification must come from the release's GitHub Actions run; local checks alone do not establish a Linux CI pass.