oh-my-opencode 5.0.0-beta.7 → 5.0.0-beta.9

package/dist/hooks/atlas/final-wave-approval-gate.test-support.d.ts

@@ -0,0 +1,50 @@
+import { mock } from "bun:test";
+import type { createAtlasHook } from "./index";
+import type { createToolExecuteAfterHandler } from "./tool-execute-after";
+import type { SessionState, ToolExecuteAfterOutput } from "./types";
+type AtlasHookContext = Parameters<typeof createAtlasHook>[0];
+type PromptMock = ReturnType<typeof mock>;
+export type FinalWaveMockPluginInput = AtlasHookContext & {
+    _promptMock: PromptMock;
+};
+export type FinalWaveAfterHandlerHarness = {
+    sessionState: SessionState;
+    run: (toolOutput: ToolExecuteAfterOutput) => Promise<void>;
+};
+/**
+ * Builds a mock plugin input with a recording prompt mock. `resolveParentSessionID`
+ * maps a subagent task session id to the orchestrator session the task belongs to.
+ */
+export declare function createFinalWaveMockPluginInput(options: {
+    directory: string;
+    resolveParentSessionID: (taskSessionID: string) => string;
+}): FinalWaveMockPluginInput;
+/**
+ * Registers per-test temp-directory lifecycle hooks (`.omo` scaffold + boulder state
+ * cleanup) and exposes the current directory. `resetAgentRegistration` also resets
+ * the shared claude-code session-state registry between tests.
+ */
+export declare function registerFinalWaveTestEnvironment(options?: {
+    resetAgentRegistration?: boolean;
+}): {
+    readonly directory: string;
+};
+/** Writes the plan markdown plus the boulder state pointing the session at it. */
+export declare function writeFinalWavePlanState(options: {
+    directory: string;
+    sessionID: string;
+    planName: string;
+    planContent: string;
+}): string;
+/**
+ * Wraps `createToolExecuteAfterHandler` with a shared in-memory session-state map so
+ * tests can assert pause/count flags. The handler factory is passed in explicitly:
+ * suites that mock modules (`mock.module`) must hand over their post-mock import so
+ * the harness runs against the same module instance the suite asserts on.
+ */
+export declare function createFinalWaveAfterHandlerHarness(options: {
+    ctx: AtlasHookContext;
+    sessionID: string;
+    createHandler: typeof createToolExecuteAfterHandler;
+}): FinalWaveAfterHandlerHarness;
+export {};

package/dist/hooks/atlas/system-reminder-templates.d.ts

@@ -2,6 +2,5 @@ export declare const DIRECT_WORK_REMINDER: string;
 export declare const BOULDER_CONTINUATION_PROMPT: string;
 export declare const BOULDER_COMPLETE_PROMPT = "<system-reminder>\nBOULDER COMPLETE: plan \"{PLAN_NAME}\" is fully checked.\n\nTotal elapsed: {ELAPSED_HUMAN}\n\nPer-task breakdown:\n{TASK_BREAKDOWN}\n\nPer your <boulder_completion_response> instructions, print the final ORCHESTRATION COMPLETE summary in your next turn. This nudge fires at most once.\n</system-reminder>";
 export declare const VERIFICATION_REMINDER = "**THE SUBAGENT JUST CLAIMED THIS TASK IS DONE. THEY ARE PROBABLY LYING.**\n\nSubagents say \"done\" when code has errors, tests pass trivially, logic is wrong,\nor they quietly added features nobody asked for. This happens EVERY TIME.\nAssume the work is broken until YOU prove otherwise.\n\n---\n\n**PHASE 1: READ THE CODE FIRST (before running anything)**\n\nDo NOT run tests yet. Read the code FIRST so you know what you're testing.\n\n1. `Bash(\"git diff --stat -- ':!node_modules'\")` - see exactly which files changed. Any file outside expected scope = scope creep.\n2. `Read` EVERY changed file - no exceptions, no skimming.\n3. For EACH file, critically ask:\n   - Does this code ACTUALLY do what the task required? (Re-read the task, compare line by line)\n   - Any stubs, TODOs, placeholders, hardcoded values? (`Grep` for TODO, FIXME, HACK, xxx)\n   - Logic errors? Trace the happy path AND the error path in your head.\n   - Anti-patterns? (`Grep` for `as any`, `@ts-ignore`, empty catch, console.log in changed files)\n   - Scope creep? Did the subagent touch things or add features NOT in the task spec?\n4. Cross-check every claim:\n   - Said \"Updated X\" - READ X. Actually updated, or just superficially touched?\n   - Said \"Added tests\" - READ the tests. Do they test REAL behavior or just `expect(true).toBe(true)`?\n   - Said \"Follows patterns\" - OPEN a reference file. Does it ACTUALLY match?\n\n**If you cannot explain what every changed line does, you have NOT reviewed it.**\n\n**PHASE 2: RUN AUTOMATED CHECKS (targeted, then broad)**\n\nNow that you understand the code, verify mechanically:\n1. `lsp_diagnostics` on EACH changed file - ZERO new errors\n2. Run tests for changed modules FIRST, then full suite\n3. Build/typecheck - exit 0\n\nIf Phase 1 found issues but Phase 2 passes: Phase 2 is WRONG. The code has bugs that tests don't cover. Fix the code.\n\n**PHASE 3: HANDS-ON QA - ACTUALLY RUN IT (MANDATORY for user-facing changes)**\n\nTests and linters CANNOT catch: visual bugs, wrong CLI output, broken user flows, API response shape issues.\n\n**If this task produced anything a user would SEE or INTERACT with, you MUST launch it and verify yourself.**\n\n- **Frontend/UI**: `/playwright` skill - load the page, click through the flow, check console. Verify: page loads, interactions work, console clean, responsive.\n- **TUI/CLI**: `interactive_bash` - run the command, try good input, try bad input, try --help. Verify: command runs, output correct, error messages helpful, edge inputs handled.\n- **API/Backend**: `Bash` with curl - hit the endpoint, check response body, send malformed input. Verify: returns 200, body correct, error cases return proper errors.\n- **Config/Build**: Actually start the service or import the config. Verify: loads without error, backward compatible.\n\nThis is NOT optional \"if applicable\". If the deliverable is user-facing and you did not run it, you are shipping untested work.\n\n**PHASE 4: GATE DECISION - Should you proceed to the next task?**\n\nAnswer honestly:\n1. Can I explain what EVERY changed line does? (If no - back to Phase 1)\n2. Did I SEE it work with my own eyes? (If user-facing and no - back to Phase 3)\n3. Am I confident nothing existing is broken? (If no - run broader tests)\n\nALL three must be YES. \"Probably\" = NO. \"I think so\" = NO. Investigate until CERTAIN.\n\n- **All 3 YES** - Proceed: mark task complete, move to next.\n- **Any NO** - Reject: resume with `task_id`, fix the specific issue.\n- **Unsure** - Reject: \"unsure\" = \"no\". Investigate until you have a definitive answer.\n\n**DO NOT proceed to the next task until all 4 phases are complete and the gate passes.**";
-export declare const VERIFICATION_REMINDER_GEMINI = "**THE SUBAGENT HAS FINISHED. THEIR WORK IS EXTREMELY SUSPICIOUS.**\n\nThe subagent CLAIMS this task is done. Based on thousands of executions, subagent claims are FALSE more often than true.\nThey ROUTINELY:\n- Ship code with syntax errors they didn't bother to check\n- Create stub implementations with TODOs and call it \"done\"\n- Write tests that pass trivially (testing nothing meaningful)\n- Implement logic that does NOT match what was requested\n- Add features nobody asked for and call it \"improvement\"\n- Report \"all tests pass\" when they didn't run any tests\n\n**This is NOT a theoretical warning. This WILL happen on this task. Assume the work is BROKEN.**\n\n**YOU MUST VERIFY WITH ACTUAL TOOL CALLS. NOT REASONING. TOOL CALLS.**\nThinking \"it looks correct\" is NOT verification. Running `lsp_diagnostics` IS.\n\n---\n\n**PHASE 1: READ THE CODE FIRST (DO NOT SKIP - DO NOT RUN TESTS YET)**\n\nRead the code FIRST so you know what you're testing.\n\n1. `Bash(\"git diff --stat -- ':!node_modules'\")` - see exactly which files changed.\n2. `Read` EVERY changed file - no exceptions, no skimming.\n3. For EACH file:\n   - Does this code ACTUALLY do what the task required? RE-READ the task spec.\n   - Any stubs, TODOs, placeholders? `Grep` for TODO, FIXME, HACK, xxx\n   - Anti-patterns? `Grep` for `as any`, `@ts-ignore`, empty catch\n   - Scope creep? Did the subagent add things NOT in the task spec?\n4. Cross-check EVERY claim against actual code.\n\n**If you cannot explain what every changed line does, GO BACK AND READ AGAIN.**\n\n**PHASE 2: RUN AUTOMATED CHECKS**\n\n1. `lsp_diagnostics` on EACH changed file - ZERO new errors. ACTUALLY RUN THIS.\n2. Run tests for changed modules, then full suite. ACTUALLY RUN THESE.\n3. Build/typecheck - exit 0.\n\nIf Phase 1 found issues but Phase 2 passes: Phase 2 is WRONG. Fix the code.\n\n**PHASE 3: HANDS-ON QA (MANDATORY for user-facing changes)**\n\n- **Frontend/UI**: `/playwright`\n- **TUI/CLI**: `interactive_bash`\n- **API/Backend**: `Bash` with curl\n\n**If user-facing and you did not run it, you are shipping UNTESTED BROKEN work.**\n\n**PHASE 4: GATE DECISION**\n\n1. Can I explain what EVERY changed line does? (If no \u2192 Phase 1)\n2. Did I SEE it work via tool calls? (If user-facing and no \u2192 Phase 3)\n3. Am I confident nothing is broken? (If no \u2192 broader tests)\n\nALL three must be YES. \"Probably\" = NO. \"I think so\" = NO.\n\n**DO NOT proceed to the next task until all 4 phases are complete.**";
 export declare const ORCHESTRATOR_DELEGATION_REQUIRED: string;
 export declare const SINGLE_TASK_DIRECTIVE: string;

package/dist/index.js

@@ -5716,7 +5716,7 @@ function resolveOmoTaskSettings(input, resolveParallelism = availableParallelism
     residency_max_children: record8["residency_max_children"] ?? Math.max(8, resolveParallelism() * 3)
   });
 }
-var ResidencyMaxChildrenInputSchema, OmoTaskWaitSchema, OmoTaskTeamSettingsSchema, OmoTaskWarningsSchema, OmoTaskSettingsSchema, OmoTaskWaitLayerSchema, OmoTaskTeamSettingsLayerSchema, OmoTaskWarningsLayerSchema, OmoTaskSettingsLayerSchema;
+var ResidencyMaxChildrenInputSchema, OmoTaskWaitSchema, OmoTaskTeamSettingsSchema, OmoTaskWarningsSchema, OmoTaskDagSettingsSchema, OmoTaskSettingsSchema, OmoTaskDagSettingsLayerSchema, OmoTaskWaitLayerSchema, OmoTaskTeamSettingsLayerSchema, OmoTaskWarningsLayerSchema, OmoTaskSettingsLayerSchema;
 var init_task = __esm(() => {
   ResidencyMaxChildrenInputSchema = z9.union([z9.number().int().positive(), z9.literal("unlimited")]);
   OmoTaskWaitSchema = z9.object({
@@ -5732,6 +5732,16 @@ var init_task = __esm(() => {
   OmoTaskWarningsSchema = z9.object({
     unavailable_categories: z9.boolean().default(true)
   }).strict();
+  OmoTaskDagSettingsSchema = z9.object({
+    max_nodes_per_run: z9.number().int().positive().default(64),
+    max_runs_per_session: z9.number().int().positive().default(16),
+    subscriber_ring: z9.number().int().positive().default(1000),
+    heartbeat_ms: z9.number().int().positive().default(15000),
+    history_default_limit: z9.number().int().positive().default(256),
+    history_max_limit: z9.number().int().positive().default(1000),
+    retention_days: z9.number().int().positive().default(7),
+    max_prompt_bytes: z9.number().int().positive().default(262144)
+  }).strict();
   OmoTaskSettingsSchema = z9.object({
     default_execution_mode: z9.enum(["in-process", "process"]).default("in-process"),
     default_concurrency: z9.number().int().positive().default(5),
@@ -5749,7 +5759,18 @@ var init_task = __esm(() => {
       max_members: 8,
       max_parallel_members: 4,
       max_wall_clock_minutes: 120
-    })
+    }),
+    dag: OmoTaskDagSettingsSchema.optional()
+  }).strict();
+  OmoTaskDagSettingsLayerSchema = z9.object({
+    max_nodes_per_run: z9.number().int().positive().optional(),
+    max_runs_per_session: z9.number().int().positive().optional(),
+    subscriber_ring: z9.number().int().positive().optional(),
+    heartbeat_ms: z9.number().int().positive().optional(),
+    history_default_limit: z9.number().int().positive().optional(),
+    history_max_limit: z9.number().int().positive().optional(),
+    retention_days: z9.number().int().positive().optional(),
+    max_prompt_bytes: z9.number().int().positive().optional()
   }).strict();
   OmoTaskWaitLayerSchema = z9.object({
     min_ms: z9.number().int().positive().optional(),
@@ -5777,7 +5798,8 @@ var init_task = __esm(() => {
     resume_children: z9.boolean().optional(),
     warnings: OmoTaskWarningsLayerSchema.optional(),
     wait: OmoTaskWaitLayerSchema.optional(),
-    team: OmoTaskTeamSettingsLayerSchema.optional()
+    team: OmoTaskTeamSettingsLayerSchema.optional(),
+    dag: OmoTaskDagSettingsLayerSchema.optional()
   }).strict();
 });
 
@@ -24504,10 +24526,10 @@ var CATEGORY_MODEL_REQUIREMENTS = {
   quick: {
     fallbackChain: [
       { providers: ["kimi-for-coding"], model: "kimi-for-coding-highspeed" },
-      { providers: ["quotio-openai"], model: "gpt-5.6-luna-fast", variant: "low" },
+      { providers: ["openai-codex"], model: "gpt-5.6-luna-fast", variant: "low" },
       { providers: ["deepseek"], model: "deepseek-v4-flash", variant: "off" },
       {
-        providers: ["qwen-token-plan", "alibaba-token-plan", "bailian-coding-plan", "opencode-go", "vercel"],
+        providers: ["qwen-token-plan", "alibaba-token-plan", "bailian-coding-plan", "vercel"],
         model: "qwen3.6-flash",
         variant: "low"
       },
@@ -24523,6 +24545,11 @@ var CATEGORY_MODEL_REQUIREMENTS = {
   },
   "unspecified-low": {
     fallbackChain: [
+      {
+        providers: ["xai", "github-copilot", "opencode", "vercel"],
+        model: "grok-4.6",
+        variant: "xhigh"
+      },
       {
         providers: ["openai", "quotio-openai", "github-copilot", "opencode", "vercel"],
         model: "gpt-5.6-terra",
@@ -24663,6 +24690,16 @@ function isGlmModel(model) {
   const modelName = extractModelName(model).toLowerCase();
   return modelName.includes("glm");
 }
+var GROK_45_RE = /grok-4-5(?![0-9])/;
+var GROK_46_RE = /grok-4-6(?![0-9])/;
+function isGrok45Model(model) {
+  const modelName = extractModelName(model).toLowerCase().replaceAll(".", "-");
+  return GROK_45_RE.test(modelName);
+}
+function isGrok46Model(model) {
+  const modelName = extractModelName(model).toLowerCase().replaceAll(".", "-");
+  return GROK_46_RE.test(modelName);
+}
 var GEMINI_PROVIDERS = ["google/", "google-vertex/"];
 function isGeminiModel(model) {
   if (GEMINI_PROVIDERS.some((prefix) => model.startsWith(prefix)))
@@ -28182,6 +28219,7 @@ init_logger2();
 
 // packages/omo-opencode/src/shared/normalize-sdk-response.ts
 function normalizeSDKResponse(response, fallback, options) {
+  const fallbackIsArray = Array.isArray(fallback);
   if (response == null) {
     return fallback;
   }
@@ -28191,14 +28229,17 @@ function normalizeSDKResponse(response, fallback, options) {
   if (typeof response === "object" && "data" in response) {
     const data = response.data;
     if (data != null) {
+      if (fallbackIsArray && !Array.isArray(data)) {
+        return fallback;
+      }
       return data;
     }
-    if (options?.preferResponseOnMissingData === true) {
+    if (options?.preferResponseOnMissingData === true && !fallbackIsArray) {
       return response;
     }
     return fallback;
   }
-  if (options?.preferResponseOnMissingData === true) {
+  if (options?.preferResponseOnMissingData === true && !fallbackIsArray) {
     return response;
   }
   return fallback;
@@ -28852,7 +28893,7 @@ This is NOT a default choice - it's for genuinely unclassifiable moderate-effort
 </Category_Context>
 
 <Caller_Warning>
-THIS CATEGORY USES A LIGHTWEIGHT MODEL (gpt-5.6-luna).
+THIS CATEGORY USES GROK 4.6 (xhigh).
 
 **PROVIDE CLEAR STRUCTURE:**
 1. MUST DO: Enumerate required actions explicitly
@@ -28882,7 +28923,7 @@ var OPENAI_CATEGORIES = [
   },
   {
     name: "unspecified-low",
-    config: { model: "openai/gpt-5.6-luna", variant: "xhigh" },
+    config: { model: "xai/grok-4.6", variant: "xhigh" },
     description: "Tasks that don't fit other categories, low effort required",
     promptAppend: UNSPECIFIED_LOW_CATEGORY_PROMPT_APPEND
   }
@@ -84199,6 +84240,47 @@ var model_capabilities_generated_default = {
         input: 272000,
         output: 128000
       }
+    },
+    "grok-4.6": {
+      id: "grok-4.6",
+      family: "grok",
+      reasoning: true,
+      temperature: true,
+      toolCall: true,
+      modalities: {
+        input: [
+          "text",
+          "image"
+        ],
+        output: [
+          "text"
+        ]
+      },
+      limit: {
+        context: 500000,
+        output: 500000
+      }
+    },
+    "xai/grok-4.6": {
+      id: "xai/grok-4.6",
+      family: "grok",
+      reasoning: true,
+      temperature: true,
+      toolCall: true,
+      modalities: {
+        input: [
+          "text",
+          "image",
+          "pdf"
+        ],
+        output: [
+          "text"
+        ]
+      },
+      limit: {
+        context: 500000,
+        output: 500000
+      }
     }
   }
 };
@@ -84334,21 +84416,21 @@ function restoreMessageCursor(sessionID, cursor) {
   sessionCursors.set(sessionID, cloneCursorState(cursor));
 }
 // packages/omo-opencode/src/shared/shell-env.ts
-function detectShellType() {
-  if (process.env.SHELL) {
-    const shell = process.env.SHELL;
+function detectShellType(platform = process.platform, env2 = process.env) {
+  if (env2.SHELL) {
+    const shell = env2.SHELL;
     if (shell.includes("csh") || shell.includes("tcsh")) {
       return "csh";
     }
     return "unix";
   }
-  if (process.platform === "win32" && (process.env.BASH_VERSION || process.env.MSYSTEM || process.env.WSL_DISTRO_NAME)) {
+  if (platform === "win32" && (env2.BASH_VERSION || env2.MSYSTEM || env2.WSL_DISTRO_NAME)) {
     return "unix";
   }
-  if (process.env.PSModulePath) {
+  if (env2.PSModulePath) {
     return "powershell";
   }
-  return process.platform === "win32" ? "cmd" : "unix";
+  return platform === "win32" ? "cmd" : "unix";
 }
 function shellEscape(value, shellType) {
   if (value === "") {
@@ -98644,7 +98726,7 @@ function getCachedVersion(options = {}) {
 // package.json
 var package_default = {
   name: "oh-my-opencode",
-  version: "5.0.0-beta.7",
+  version: "5.0.0-beta.9",
   description: "The Best AI Agent Harness - Batteries-Included OpenCode Plugin with Multi-Model Orchestration, Parallel Background Agents, and Crafted LSP/AST Tools",
   main: "./dist/index.js",
   types: "dist/index.d.ts",
@@ -98780,6 +98862,7 @@ var package_default = {
     "typecheck:script": "tsgo --noEmit -p script/tsconfig.json",
     test: "bun test",
     "test:codex": "bun run build:codex-install && bun run build:git-bash-mcp && bun run build:lsp-tools-mcp && bun run build:lsp-daemon && npm --prefix packages/lsp-tools-mcp test && npm --prefix packages/omo-codex/plugin ci && npm --prefix packages/omo-codex/plugin/components/ulw-loop test && bun run --cwd packages/omo-codex/plugin build && npm --prefix packages/omo-codex/plugin/components/codegraph run typecheck && npm --prefix packages/omo-codex/plugin/components/codegraph test && node scripts/check-third-party-notices.mjs --ship && bun test packages/omo-opencode/src/cli/cli-installer.platform.test.ts packages/omo-codex/src/install/codex-cache.test.ts packages/omo-codex/src/install/codex-cleanup.test.ts packages/omo-codex/src/install/codex-config-agent-cleanup.test.ts packages/omo-codex/src/install/codex-config-autonomous-features.test.ts packages/omo-codex/src/install/codex-config-reasoning.test.ts packages/omo-codex/src/install/codex-config-toml.test.ts packages/omo-codex/src/install/codex-project-local-cleanup.test.ts packages/omo-codex/src/install/install-codex-project-local-cleanup.test.ts packages/omo-codex/src/install/install-codex.test.ts packages/omo-codex/src/install/install-codex-packaged.test.ts packages/omo-codex/src/install/link-cached-plugin-agents.test.ts packages/omo-codex/src/**/*.test.ts packages/utils/src/jsonc-parser.test.ts packages/utils/src/frontmatter.test.ts packages/hashline-core/src/hash-computation.test.ts packages/hashline-core/src/smoke-untested-modules.test.ts packages/rules-engine/src/index.test.ts packages/rules-engine/src/security-boundary.test.ts packages/agents-md-core/src/injector.test.ts packages/omo-codex/plugin/components/lsp/test/package-smoke.test.ts && node --test packages/omo-codex/plugin/test/*.test.mjs packages/omo-codex/scripts/install-cache-copy.test.mjs packages/omo-codex/scripts/install-cli-args.test.mjs packages/omo-codex/scripts/install-delegated-command.test.mjs packages/omo-codex/scripts/install-config-autonomous-features.test.mjs packages/omo-codex/scripts/install-config-autonomous.test.mjs packages/omo-codex/scripts/install-config-reasoning.test.mjs packages/omo-codex/scripts/install-config.test.mjs packages/omo-codex/scripts/install-hook-targets.test.mjs packages/omo-codex/scripts/install-project-local-cleanup.test.mjs packages/omo-codex/scripts/install-lazycodex-version-stamp.test.mjs packages/omo-codex/scripts/install-local-entrypoint.test.mjs packages/omo-codex/scripts/install-local-git-bash-preflight.test.mjs packages/omo-codex/scripts/install-local.test.mjs packages/omo-codex/scripts/install-marketplace-cache.test.mjs packages/omo-codex/scripts/install-mcp-context7-runtime.test.mjs packages/omo-codex/scripts/install-mcp-runtime.test.mjs packages/omo-codex/scripts/install-packaged-local.test.mjs packages/omo-codex/scripts/install-generated-bundle.test.mjs packages/omo-codex/scripts/install-agent-links.test.mjs packages/omo-codex/scripts/install-bin-links.test.mjs",
+    "test:fast": "bun run script/test-fast.ts",
     "test:senpi": "bun run build:senpi-plugin && tsgo --noEmit -p packages/omo-senpi/tsconfig.json && bun test packages/omo-senpi",
     "test:windows-codex": "bun run test:codex",
     "build:git-bash-mcp": "bun run --cwd packages/git-bash-mcp build",
@@ -98824,7 +98907,7 @@ var package_default = {
     zod: "^4.4.3"
   },
   devDependencies: {
-    "@code-yeongyu/senpi": "2026.8.12-4",
+    "@code-yeongyu/senpi": "2026.8.17",
     "@oh-my-opencode/agents-md-core": "workspace:*",
     "@oh-my-opencode/ast-grep-mcp": "workspace:*",
     "@oh-my-opencode/boulder-state": "workspace:*",
@@ -98856,7 +98939,7 @@ var package_default = {
     "@oh-my-opencode/utils": "workspace:*",
     "@types/js-yaml": "^4.0.9",
     "@types/picomatch": "^4.0.3",
-    "@typescript/native-preview": "7.0.0-dev.20260518.1",
+    "@typescript/native-preview": "7.0.0-dev.20260707.2",
     "@xterm/addon-unicode11": "^0.9.0",
     "@xterm/xterm": "^6.0.0",
     "bun-types": "1.3.14",
@@ -98865,24 +98948,24 @@ var package_default = {
     typescript: "^7.0.2"
   },
   optionalDependencies: {
-    "oh-my-opencode-darwin-arm64": "5.0.0-beta.7",
-    "oh-my-opencode-darwin-x64": "5.0.0-beta.7",
-    "oh-my-opencode-darwin-x64-baseline": "5.0.0-beta.7",
-    "oh-my-opencode-linux-arm64": "5.0.0-beta.7",
-    "oh-my-opencode-linux-arm64-musl": "5.0.0-beta.7",
-    "oh-my-opencode-linux-x64": "5.0.0-beta.7",
-    "oh-my-opencode-linux-x64-baseline": "5.0.0-beta.7",
-    "oh-my-opencode-linux-x64-musl": "5.0.0-beta.7",
-    "oh-my-opencode-linux-x64-musl-baseline": "5.0.0-beta.7",
-    "oh-my-opencode-windows-arm64": "5.0.0-beta.7",
-    "oh-my-opencode-windows-x64": "5.0.0-beta.7",
-    "oh-my-opencode-windows-x64-baseline": "5.0.0-beta.7"
+    "oh-my-opencode-darwin-arm64": "5.0.0-beta.9",
+    "oh-my-opencode-darwin-x64": "5.0.0-beta.9",
+    "oh-my-opencode-darwin-x64-baseline": "5.0.0-beta.9",
+    "oh-my-opencode-linux-arm64": "5.0.0-beta.9",
+    "oh-my-opencode-linux-arm64-musl": "5.0.0-beta.9",
+    "oh-my-opencode-linux-x64": "5.0.0-beta.9",
+    "oh-my-opencode-linux-x64-baseline": "5.0.0-beta.9",
+    "oh-my-opencode-linux-x64-musl": "5.0.0-beta.9",
+    "oh-my-opencode-linux-x64-musl-baseline": "5.0.0-beta.9",
+    "oh-my-opencode-windows-arm64": "5.0.0-beta.9",
+    "oh-my-opencode-windows-x64": "5.0.0-beta.9",
+    "oh-my-opencode-windows-x64-baseline": "5.0.0-beta.9"
   },
   overrides: {
-    "@earendil-works/pi-agent-core": "0.80.3",
-    "@earendil-works/pi-ai": "0.80.3",
-    "@earendil-works/pi-coding-agent": "0.80.3",
-    "@earendil-works/pi-tui": "0.80.3",
+    "@earendil-works/pi-agent-core": "0.84.2",
+    "@earendil-works/pi-ai": "0.84.2",
+    "@earendil-works/pi-coding-agent": "0.84.2",
+    "@earendil-works/pi-tui": "0.84.2",
     hono: "^4.12.18",
     "@hono/node-server": "^1.19.13",
     "express-rate-limit": "^8.5.1",
@@ -116331,6 +116414,10 @@ async function handleAtlasSessionIdle(input) {
     return;
   }
   const { boulderState, progress, appendedSession } = activeBoulderSession;
+  if (sessionState.waitingForFinalWaveApproval) {
+    log2(`[${HOOK_NAME9}] Skipped: waiting for explicit final-wave approval`, { sessionID });
+    return;
+  }
   if (progress.isComplete) {
     await handleCompletedBoulderIdle({ ctx, options, sessionID, sessionState, boulderState });
     return;
@@ -116358,10 +116445,6 @@ async function handleAtlasSessionIdle(input) {
   const now = Date.now();
   const activePlanPath = resolveBoulderPlanPath(ctx.directory, boulderState);
   resetStallStateForPlanChange(sessionState, activePlanPath);
-  if (sessionState.waitingForFinalWaveApproval) {
-    log2(`[${HOOK_NAME9}] Skipped: waiting for explicit final-wave approval`, { sessionID });
-    return;
-  }
   if (sessionState.stalledContinuationReason) {
     log2(`[${HOOK_NAME9}] Skipped: boulder continuation stalled`, {
       sessionID,
@@ -152907,6 +152990,12 @@ function buildGptSisyphusAgentConfig(mode, model, prompt) {
 function buildGlmSisyphusAgentConfig(mode, model, prompt) {
   return buildBaseSisyphusAgentConfig(mode, model, prompt);
 }
+function buildGrokSisyphusAgentConfig(mode, model, prompt) {
+  return {
+    ...buildBaseSisyphusAgentConfig(mode, model, prompt),
+    reasoningEffort: "high"
+  };
+}
 function buildClaudeSisyphusAgentConfig(mode, model, prompt) {
   return {
     ...buildBaseSisyphusAgentConfig(mode, model, prompt),
@@ -156276,6 +156365,158 @@ function buildGpt55SisyphusPrompt(model, availableAgents, _availableTools = [],
 ${body}`;
 }
 
+// packages/omo-opencode/src/agents/sisyphus/grok-4.ts
+function getGrokPromptIdentity(model) {
+  return isGrok46Model(model) ? "Grok 4.6" : "Grok 4.5";
+}
+function buildGrok4TasksSection(useTaskSystem) {
+  const noun = useTaskSystem ? "tasks" : "todos";
+  const create = useTaskSystem ? "task_create" : "todowrite";
+  const update = useTaskSystem ? "task_update" : "todowrite";
+  const hook = useTaskSystem ? "TASK CONTINUATION" : "TODO CONTINUATION";
+  return `<tasks>
+Use ${noun} for implementation work with two or more real steps, cross-file edits, delegated work, or uncertain scope. Skip tracking for direct answers, pure exploration, and one-step edits.
+
+When tracking: call \`${create}\` before implementation, keep exactly one item \`in_progress\`, and call \`${update}\` the moment an item is done. Never batch completions. If scope changes, revise the list before more edits.
+
+Your ${noun} are tracked by the harness via [SYSTEM REMINDER - ${hook}].
+</tasks>`;
+}
+function buildGrok4SisyphusPrompt(model, availableAgents, availableTools = [], availableSkills = [], availableCategories = [], useTaskSystem = false) {
+  const keyTriggers = buildKeyTriggersSection(availableAgents, availableSkills);
+  const toolSelection = buildToolSelectionTable(availableAgents, availableTools, availableSkills);
+  const exploreSection = buildExploreSection(availableAgents);
+  const librarianSection = buildLibrarianSection(availableAgents);
+  const categorySkillsGuide = buildCategorySkillsDelegationGuide(availableCategories, availableSkills);
+  const delegationTable = buildDelegationTable(availableAgents);
+  const oracleSection = buildOracleSection(availableAgents);
+  const hardBlocks = buildHardBlocksSection();
+  const antiPatterns = buildAntiPatternsSection();
+  const nonClaudePlannerSection = buildNonClaudePlannerSection(model);
+  const tasksSection = buildGrok4TasksSection(useTaskSystem);
+  const agentIdentity = buildAgentIdentitySection("Sisyphus", "Powerful AI Agent with orchestration capabilities from OhMyOpenCode");
+  const roleBlock = `<role>
+You are Sisyphus, the OhMyOpenCode orchestration lead, running on ${getGrokPromptIdentity(model)}.
+
+You are a senior engineer who scales output through specialists. Understand the user's destination, route the work to the right specialist, verify with real evidence, and stop only when the result is production ready.
+
+Implementation starts only when the current user turn explicitly asks for it with concrete scope. Questions get answers, investigations get findings, implementation requests get shipped work.
+</role>`;
+  const calibrationBlock = `<grok_calibration>
+Your judgment is good; this prompt stays short on purpose and trusts you to fill gaps with taste. Four rules carry the leverage:
+
+1. DONE IS WRITTEN DOWN. Before implementation, state what done means in one line: observable acceptance criteria, not a vibe. You verify against exactly those criteria, and you neither stop short of them nor expand past them.
+2. VERIFY, THEN ITERATE. Verify the function and the design after implementation, and keep iterating and verifying until it is production ready. One pass of "it runs" is not done.
+3. CAPTURE, LIST, FIX. When output is hard to inspect (visual, layout, motion, formatted documents), capture the current state, list concretely what is wrong with it, then fix only those things. Never "improve" blind.
+4. NO REPEATED BLOCKS. You tend to duplicate code across components. The second time a block appears, extract and share it instead of pasting a third copy.
+</grok_calibration>`;
+  const intentBlock = `<intent>
+Classify the CURRENT user message only. Do not carry implementation authorization across turns.
+
+${keyTriggers}
+
+Surface form to routing:
+
+| User says | True intent | You do |
+|---|---|---|
+| "explain", "how does" | understanding | explore enough, then answer |
+| "implement", "add", "create", "write" | implementation | plan, delegate or execute, verify |
+| "look into", "check", "investigate" | investigation | inspect, report findings, wait |
+| "what do you think" | evaluation | judge, propose, wait |
+| "broken", "error", "fix" | root-cause repair | diagnose, fix minimally, verify |
+| "refactor", "improve", "clean up" | open-ended change | assess, propose or use the matching skill |
+
+Say one concise intent line before non-trivial action: "I read this as [type]: [route]." If the answer is already in context, answer instead of re-deriving.
+
+Ask only for scope changes, critical missing information, destructive actions, or external side effects. Minor decisions (names, defaults, equivalent approaches) are yours; note the choice later.
+</intent>`;
+  const explorationBlock = `<exploration>
+Use tools for facts. Internal memory is not evidence for file contents, configs, APIs, or current project state.
+
+${toolSelection}
+
+${exploreSection}
+
+${librarianSection}
+
+Parallelize independent calls: file reads, searches, diagnostics, and background agents go out together. Sequence only when a later call needs an earlier result.
+
+Search budget: known file or symbol = direct read/search; unfamiliar local pattern = one parallel wave; external package or API = librarian; architectural risk = Oracle. Stop when sources converge, the target file set is known, or the answer is found.
+
+Fire explore/librarian in the background with [CONTEXT], [GOAL], [DOWNSTREAM], and [REQUEST]. Continue only with non-overlapping work; otherwise end the turn and wait for the completion reminder before calling \`background_output(task_id="bg_...")\`. Use \`task(task_id="ses_...")\` only for follow-ups to the same subagent.
+
+${buildAntiDuplicationSection()}
+</exploration>`;
+  const delegationBlock = `<delegation>
+Prefer delegation when a specialist fits, the work spans multiple files, the domain is visual/frontend/security/performance, or the module is unfamiliar. Execute directly only for small, local, fully understood changes.
+
+${categorySkillsGuide}
+
+${nonClaudePlannerSection}
+
+${delegationTable}
+
+Every delegation prompt carries six sections: TASK, EXPECTED OUTCOME, REQUIRED TOOLS, MUST DO, MUST NOT DO, CONTEXT. The EXPECTED OUTCOME is the delegate's definition of done - make it observable.
+
+After delegation, verify the files and behavior yourself. A subagent report is a lead, not evidence.
+${oracleSection ? `
+${oracleSection}
+` : ""}</delegation>`;
+  const executionBlock = `<behavior>
+Implementation loop:
+
+1. Write down what done means, then plan the smallest path to it. Two or more steps need ${useTaskSystem ? "tasks" : "todos"}; one obvious edit does not.
+2. Match the repo: read configs and similar files before writing. Do not invent style.
+3. Change only what the request requires. Bug fix does not mean refactor. Refactor does not mean feature work.
+4. Use type-safe code. No type suppression, no speculative fallbacks, no helpers for one-off operations, no validation away from trust boundaries.
+5. On failure, read the error, identify the root cause, try a materially different approach, and re-verify. After three failed approaches, stop editing and consult Oracle, or ask if Oracle cannot resolve it.
+
+Never revert, delete, push, publish, message, or affect shared systems without explicit approval. Reversible local edits and verification commands are allowed.
+</behavior>`;
+  const verificationBlock = `<verification>
+Verification defines done, and it loops until production ready.
+
+- File edit: run \`lsp_diagnostics\` on every changed file.
+- Behavioral change: run adjacent tests or the smallest relevant suite.
+- Buildable project: run the build/typecheck path that covers the touched code.
+- Runnable or user-visible behavior: exercise the real surface - browser for web, interactive_bash for TUI/CLI, curl for HTTP, driver script for libraries. Click through the real user path, not just the happy entry point.
+- Hard-to-inspect output: capture the current state, list what is wrong, fix only those things, capture again.
+- Delegated work: inspect touched files and rerun checks yourself.
+
+A pass that surfaces a defect goes back to step one of the loop, not into the report. Report only evidence from this turn: "should pass" means unverified. Fix failures caused by your change; name unrelated pre-existing failures without widening scope.
+</verification>`;
+  const communicationBlock = `<communication>
+Every sentence carries information the user does not already have. Never restate the task back, never narrate routine tool calls, no flattery or filler.
+
+Stay quiet through small changes; start narrating when you touch many files or change direction. Final answers state what changed, where, the verification evidence, and any real residual risk - dense and short.
+</communication>`;
+  const constraintsBlock = `<constraints>
+${hardBlocks}
+
+${antiPatterns}
+</constraints>`;
+  return `${agentIdentity}
+${roleBlock}
+
+${calibrationBlock}
+
+${intentBlock}
+
+${explorationBlock}
+
+${delegationBlock}
+
+${executionBlock}
+
+${verificationBlock}
+
+${tasksSection}
+
+${communicationBlock}
+
+${constraintsBlock}`;
+}
+
 // packages/omo-opencode/src/agents/kimi-tool-loop-guard.ts
 var KIMI_TOOL_LOOP_GUARD = `<tool_loop_guard>
 Never call the same tool with the same arguments more than twice in a row.
@@ -157143,6 +157384,8 @@ function resolveSisyphusPromptFamily(model) {
     return "claude-opus-4-7";
   if (isGlmModel(model))
     return "glm-5-2";
+  if (isGrok45Model(model) || isGrok46Model(model))
+    return "grok-4";
   return "fallback";
 }
 function createSisyphusAgent(model, availableAgents, availableToolNames, availableSkills, availableCategories, useTaskSystem = false) {
@@ -157171,6 +157414,8 @@ function createSisyphusAgent(model, availableAgents, availableToolNames, availab
       return buildClaudeSisyphusAgentConfig(MODE, model, buildClaudeOpus47SisyphusPrompt(model, agents, tools, skills2, categories2, useTaskSystem));
     case "glm-5-2":
       return buildGlmSisyphusAgentConfig(MODE, model, buildGlm52SisyphusPrompt(model, agents, tools, skills2, categories2, useTaskSystem));
+    case "grok-4":
+      return buildGrokSisyphusAgentConfig(MODE, model, buildGrok4SisyphusPrompt(model, agents, tools, skills2, categories2, useTaskSystem));
     case "fallback": {
       const prompt = buildFallbackSisyphusPrompt(model, agents, tools, skills2, categories2, useTaskSystem);
       return isGptModel(model) ? buildGptSisyphusAgentConfig(MODE, model, prompt) : buildClaudeSisyphusAgentConfig(MODE, model, prompt);

package/dist/shared/normalize-sdk-response.d.ts

@@ -1,4 +1,5 @@
 export interface NormalizeSDKResponseOptions {
+    /** Return the raw response when data is missing, except when the fallback requires an array. */
     preferResponseOnMissingData?: boolean;
 }
 export declare function normalizeSDKResponse<TData>(response: unknown, fallback: TData, options?: NormalizeSDKResponseOptions): TData;

package/dist/shared/shell-env.d.ts

@@ -13,7 +13,7 @@ export type ShellType = "unix" | "powershell" | "cmd" | "csh";
  * specific (BASH_VERSION, MSYSTEM, WSL_DISTRO_NAME) — TERM is excluded
  * because some PowerShell users set it manually.
  */
-export declare function detectShellType(): ShellType;
+export declare function detectShellType(platform?: NodeJS.Platform, env?: NodeJS.ProcessEnv): ShellType;
 /**
  * Shell-escape a value for use in environment variable assignment.
  *