claudecode-omc 4.4.4 → 4.4.6

package/docs/CLAUDE.md

@@ -1,5 +1,5 @@
 <!-- OMC:START -->
-<!-- OMC:VERSION:4.3.4 -->
+<!-- OMC:VERSION:4.5.1 -->
 # oh-my-claudecode - Intelligent Multi-Agent Orchestration
 
 You are running with oh-my-claudecode (OMC), a multi-agent orchestration layer for Claude Code.
@@ -9,7 +9,7 @@ Your role is to coordinate specialized agents, tools, and skills so work is comp
 - Delegate specialized or tool-heavy work to the most appropriate agent.
 - Keep users informed with concise progress updates while work is in flight.
 - Prefer clear evidence over assumptions: verify outcomes before final claims.
-- Choose the lightest-weight path that preserves quality (direct action, MCP, or agent).
+- Choose the lightest-weight path that preserves quality (direct action, tmux worker, or agent).
 - Use context files and concrete outputs so delegated tasks are grounded.
 - Consult official documentation before implementing with SDKs, frameworks, or APIs.
 </operating_principles>
@@ -27,7 +27,7 @@ Work directly only for trivial operations where delegation adds disproportionate
 
 For substantive code changes, route implementation to `executor` (or `deep-executor` for complex autonomous execution). This keeps editing workflows consistent and easier to verify.
 
-For non-trivial or uncertain SDK/API/framework usage, delegate to `document-specialist` to fetch official docs first. Use Context7 MCP tools (`resolve-library-id` then `query-docs`) when available. This prevents guessing field names or API contracts. For well-known, stable APIs you can proceed directly.
+For non-trivial or uncertain SDK/API/framework usage, delegate to `document-specialist` to fetch official docs first. This prevents guessing field names or API contracts. For well-known, stable APIs you can proceed directly.
 </delegation_rules>
 
 <model_routing>
@@ -88,45 +88,11 @@ Compatibility aliases may still be normalized during routing, but canonical runt
 
 ---
 
-<mcp_routing>
-For read-only analysis tasks, prefer MCP tools over spawning Claude agents -- they are faster and cheaper.
-
-**IMPORTANT -- Deferred Tool Discovery:** MCP tools (`ask_codex`, `ask_gemini`, and their job management tools) are deferred and NOT in your tool list at session start. Use this 3-step discovery sequence before your first MCP tool use:
-1. `ToolSearch("mcp")` -- broad search; finds all deferred MCP tools at once
-2. From results, select the full tool name: look for `mcp__x__ask_codex` or `mcp__g__ask_gemini`
-3. Fall back to the equivalent Claude agent **only if** step 1 returns no results
-
-**Never use `ToolSearch("ask_codex")` or `ToolSearch("ask_gemini")` as the primary search** -- narrow searches can return false negatives even when MCP tools are present. Always start with `ToolSearch("mcp")`.
-If ToolSearch returns no results, the MCP server is not configured -- fall back to the equivalent Claude agent. Never block on unavailable MCP tools.
-
-Available MCP providers:
-- Codex (`mcp__x__ask_codex`): OpenAI gpt-5.3-codex -- code analysis, planning validation, review
-- Gemini (`mcp__g__ask_gemini`): Google gemini-3-pro-preview -- design across many files (1M context)
-
-Any OMC agent role can be passed as `agent_role` to either provider. The role loads a matching system prompt if one exists; otherwise the task runs without role-specific framing.
-
-Provider strengths (use these to choose the right provider):
-- **Codex excels at**: architecture review, planning validation, critical analysis, code review, security review, test strategy. Recommended roles: architect, planner, critic, analyst, code-reviewer, security-reviewer, test-engineer.
-- **Gemini excels at**: UI/UX design review, documentation, visual analysis, large-context tasks (1M tokens). Recommended roles: designer, writer.
-
-Always attach `context_files`/`files` when calling MCP tools. MCP output is advisory -- verification (tests, typecheck) should come from tool-using agents.
-
-Background pattern: spawn with `background: true`, check with `check_job_status`, await with `wait_for_job` (up to 1 hour).
-
-Agents that have no MCP replacement (they need Claude's tool access): `executor`, `deep-executor`, `explore`, `debugger`, `verifier`, `scientist`, `build-fixer`, `qa-tester`, all review-lane agents.
-
-Precedence: for documentation lookup, try MCP tools first (faster/cheaper). For synthesis, evaluation, or implementation guidance on external packages, use `document-specialist`.
-
-MCP output is wrapped as untrusted content; response files have output safety constraints applied.
-</mcp_routing>
-
----
-
 <tools>
-External AI (MCP providers):
-- Codex: `mcp__x__ask_codex` with `agent_role` (any role; best for: architect, planner, critic, analyst, code-reviewer, security-reviewer, test-engineer)
-- Gemini: `mcp__g__ask_gemini` with `agent_role` (any role; best for: designer, writer)
-- Job management: `check_job_status`, `wait_for_job`, `kill_job`, `list_jobs` (per provider)
+External AI (tmux CLI workers):
+- For **Claude agents**: use `/team N:executor "task"` — spawns Claude Code agent teammates via `TeamCreate`/`Task`
+- For **Codex or Gemini CLI workers**: use `/omc-teams N:codex "task"` or `/omc-teams N:gemini "task"` — spawns CLI processes in tmux panes via `bridge/runtime-cli.cjs`
+- omc-teams MCP tools: `mcp__team__omc_run_team_start`, `mcp__team__omc_run_team_wait`, `mcp__team__omc_run_team_status`, `mcp__team__omc_run_team_cleanup`
 
 OMC State:
 - `state_read`, `state_write`, `state_clear`, `state_list_active`, `state_get_status`
@@ -206,17 +172,11 @@ Agent Shortcuts (thin wrappers; call the agent directly with `model` for more co
 - `security-review` -> `security-reviewer`: "security review"
 - `review` -> `plan --review`: "review plan", "critique plan"
 
-MCP Delegation (auto-detected when an intent phrase is present):
-- `ask codex`, `use codex`, `delegate to codex` -> `ask_codex`
-- `ask gpt`, `use gpt`, `delegate to gpt` -> `ask_codex`
-- `ask gemini`, `use gemini`, `delegate to gemini` -> `ask_gemini`
-- Bare keywords without an intent phrase do not trigger delegation.
-
 Notifications: `configure-notifications` ("configure discord", "setup discord", "discord webhook", "configure telegram", "setup telegram", "telegram bot", "configure slack", "setup slack")
 
 Utilities: `cancel`, `note`, `learner`, `omc-setup`, `mcp-setup`, `hud`, `omc-doctor`, `omc-help`, `trace`, `release`, `project-session-manager` (`psm` is deprecated alias), `skill`, `writer-memory`, `ralph-init`, `learn-about-omc`
 
-Conflict resolution: explicit mode keywords (`ulw`, `ultrawork`) override defaults. Generic "fast"/"parallel" reads `~/.claude/.omc-config.json` -> `defaultExecutionMode`. Ralph includes ultrawork (persistence wrapper). Autopilot can transition to ralph or ultraqa. Autopilot and ultrapilot are mutually exclusive.
+Conflict resolution: explicit mode keywords (`ulw`, `ultrawork`) override defaults. Generic "fast"/"parallel" reads `~/.claude/.omc-config.json` -> `defaultExecutionMode`. Ralph includes ultrawork (persistence wrapper). Autopilot can transition to ralph or ultraqa. Autopilot and ultrapilot are mutually exclusive. Keyword disambiguation: bare "codex" or "gemini" routes to `omc-teams`; the full phrase "claude codex gemini" routes to `ccg` (longest-match priority).
 </skills>
 
 ---

package/docs/MIGRATION.md

@@ -6,6 +6,7 @@ This guide covers all migration paths for oh-my-claudecode. Find your current ve
 
 ## Table of Contents
 
+- [Unreleased: Team MCP timeoutSeconds Removal](#unreleased-team-mcp-timeoutseconds-removal)
 - [v3.5.3 → v3.5.5: Test Fixes & Cleanup](#v353--v355-test-fixes--cleanup)
 - [v3.5.2 → v3.5.3: Skill Consolidation](#v352--v353-skill-consolidation)
 - [v2.x → v3.0: Package Rename & Auto-Activation](#v2x--v30-package-rename--auto-activation)
@@ -14,6 +15,119 @@ This guide covers all migration paths for oh-my-claudecode. Find your current ve
 
 ---
 
+## Unreleased: Team MCP timeoutSeconds Removal
+
+### TL;DR
+
+`omc_run_team_start` no longer accepts `timeoutSeconds`. Passing it now returns an API error. Use `omc_run_team_wait` with `timeout_ms` to limit how long you block (workers keep running), and call `omc_run_team_cleanup` only when you explicitly want to stop worker panes.
+
+### What `timeoutSeconds` Did Before
+
+In earlier versions, `omc_run_team_start` accepted a `timeoutSeconds` parameter that would automatically kill all worker panes if the team had not completed within that duration. Example of the old pattern:
+
+```js
+// OLD — no longer works
+mcp__team__omc_run_team_start({
+  teamName: "my-team",
+  agentTypes: ["claude", "codex"],
+  tasks: [...],
+  cwd: "/path/to/project",
+  timeoutSeconds: 120   // <-- would silently kill workers after 2 minutes
+})
+```
+
+### Why It Was Removed
+
+`timeoutSeconds` was a footgun: it silently killed worker panes mid-task whenever the wall-clock limit was reached, with no way for the caller to distinguish "timed out and killed" from "completed normally". This caused data loss (workers writing files when killed), confusing error states, and made it impossible to resume or inspect stalled teams. Callers who set a conservative timeout often lost work that was 90% done.
+
+The fix separates concerns cleanly:
+- **Wait timeout** (`omc_run_team_wait timeout_ms`) — limits how long *your call* blocks. Workers keep running.
+- **Explicit cleanup** (`omc_run_team_cleanup`) — kills worker panes only when you decide to cancel.
+
+### How to Migrate
+
+**Step 1** — Remove `timeoutSeconds` from every `omc_run_team_start` call.
+
+**Step 2** — Use `omc_run_team_wait` with `timeout_ms` to bound the blocking call. If it times out, workers are still alive; call `omc_run_team_wait` again to keep waiting:
+
+```js
+// NEW — start (no timeoutSeconds)
+const { jobId } = await mcp__team__omc_run_team_start({
+  teamName: "my-team",
+  agentTypes: ["claude", "codex"],
+  tasks: [...],
+  cwd: "/path/to/project",
+})
+
+// Wait up to 5 minutes (workers keep running if this times out)
+const result = await mcp__team__omc_run_team_wait({
+  job_id: jobId,
+  timeout_ms: 300000,
+})
+
+// If result.status === 'running' the wait timed out; call wait again or inspect progress:
+// await mcp__team__omc_run_team_status({ job_id: jobId })
+// await mcp__team__omc_run_team_wait({ job_id: jobId, timeout_ms: 600000 })
+```
+
+**Step 3** — Call `omc_run_team_cleanup` only when you *intentionally* want to cancel and stop worker panes:
+
+```js
+// Explicit cancel — only when you want to stop the workers
+await mcp__team__omc_run_team_cleanup({ job_id: jobId })
+```
+
+### Common Patterns with `/omc-teams`
+
+**Pattern A — Wait with re-try on timeout (recommended)**
+
+```js
+// Start
+const { jobId } = await mcp__team__omc_run_team_start({ ... })
+
+// Wait loop: keep waiting until terminal state
+let result
+do {
+  result = await mcp__team__omc_run_team_wait({ job_id: jobId, timeout_ms: 300000 })
+} while (result.status === 'running')
+```
+
+**Pattern B — Non-blocking progress check**
+
+```js
+const { jobId } = await mcp__team__omc_run_team_start({ ... })
+
+// Do other work, then check progress
+const status = await mcp__team__omc_run_team_status({ job_id: jobId })
+if (status.status === 'running') {
+  // Not done yet; wait or come back later
+}
+```
+
+**Pattern C — Explicit cancel when you give up**
+
+```js
+const { jobId } = await mcp__team__omc_run_team_start({ ... })
+
+const result = await mcp__team__omc_run_team_wait({ job_id: jobId, timeout_ms: 60000 })
+if (result.status === 'running') {
+  // Decided to cancel — explicitly stop worker panes
+  await mcp__team__omc_run_team_cleanup({ job_id: jobId })
+}
+```
+
+### Breaking Change
+
+Passing `timeoutSeconds` to `omc_run_team_start` now returns an API error immediately:
+
+```
+Error: omc_run_team_start no longer accepts timeoutSeconds. Remove timeoutSeconds
+and use omc_run_team_wait timeout_ms to limit the wait call only (workers keep running
+until completion or explicit omc_run_team_cleanup).
+```
+
+---
+
 ## v3.5.3 → v3.5.5: Test Fixes & Cleanup
 
 ### TL;DR

package/docs/REFERENCE.md

@@ -13,7 +13,6 @@ Complete reference for oh-my-claudecode. For quick start, see the main [README.m
 - [Slash Commands](#slash-commands)
 - [Hooks System](#hooks-system)
 - [Magic Keywords](#magic-keywords)
-- [MCP Path Boundary Rules](#mcp-path-boundary-rules)
 - [Platform Support](#platform-support)
 - [Performance Monitoring](#performance-monitoring)
 - [Troubleshooting](#troubleshooting)
@@ -55,7 +54,7 @@ This integrates directly with Claude Code's plugin system and uses Node.js hooks
 Configure omc for the current project only:
 
 ```
-/oh-my-claudecode:omc-setup
+/oh-my-claudecode:omc-setup --local
 ```
 
 - Creates `./.claude/CLAUDE.md` in your current project
@@ -93,6 +92,31 @@ If both configurations exist, **project-scoped takes precedence** over global:
 ./.claude/CLAUDE.md  (project)   →  Overrides  →  ~/.claude/CLAUDE.md  (global)
 ```
 
+### Environment Variables
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `OMC_STATE_DIR` | _(unset)_ | Centralized state directory. When set, OMC stores state at `$OMC_STATE_DIR/{project-id}/` instead of `{worktree}/.omc/`. This preserves state across worktree deletions. The project identifier is derived from the git remote URL (or worktree path for local-only repos). |
+| `OMC_BRIDGE_SCRIPT` | _(auto-detected)_ | Path to the Python bridge script |
+| `OMC_PARALLEL_EXECUTION` | `true` | Enable/disable parallel agent execution |
+| `OMC_CODEX_DEFAULT_MODEL` | _(provider default)_ | Default model for Codex CLI workers |
+| `OMC_GEMINI_DEFAULT_MODEL` | _(provider default)_ | Default model for Gemini CLI workers |
+| `DISABLE_OMC` | _(unset)_ | Set to any value to disable all OMC hooks |
+| `OMC_SKIP_HOOKS` | _(unset)_ | Comma-separated list of hook names to skip |
+
+#### Centralized State with `OMC_STATE_DIR`
+
+By default, OMC stores state in `{worktree}/.omc/`. This is lost when worktrees are deleted. To preserve state across worktree lifecycles, set `OMC_STATE_DIR`:
+
+```bash
+# In your shell profile (~/.bashrc, ~/.zshrc, etc.)
+export OMC_STATE_DIR="$HOME/.claude/omc"
+```
+
+This resolves to `~/.claude/omc/{project-identifier}/` where the project identifier uses a hash of the git remote URL (stable across worktrees/clones) with a fallback to the directory path hash for local-only repos.
+
+If both a legacy `{worktree}/.omc/` directory and a centralized directory exist, OMC logs a notice and uses the centralized directory. You can then migrate data from the legacy directory and remove it.
+
 ### When to Re-run Setup
 
 - **First time**: Run after installation (choose project or global)
@@ -237,8 +261,8 @@ Always use `oh-my-claudecode:` prefix when calling via Task tool.
 | `ralph` | Self-referential development until completion | `/oh-my-claudecode:ralph` |
 | `ralph-init` | Initialize PRD for structured task tracking | `/oh-my-claudecode:ralph-init` |
 | `ultraqa` | Autonomous QA cycling workflow | `/oh-my-claudecode:ultraqa` |
-| `plan` | Start planning session | `/oh-my-claudecode:plan` |
-| `ralplan` | Iterative planning (Planner+Architect+Critic) | `/oh-my-claudecode:ralplan` |
+| `plan` | Start planning session (consensus mode uses RALPLAN-DR structured deliberation) | `/oh-my-claudecode:plan` |
+| `ralplan` | Iterative planning (Planner+Architect+Critic) with structured deliberation; short mode default, `--deliberate` for high-risk pre-mortem + expanded test plan | `/oh-my-claudecode:ralplan` |
 | `review` | Review work plans with critic | `/oh-my-claudecode:review` |
 
 ### Enhancement Skills
@@ -292,8 +316,8 @@ All skills are available as slash commands with the prefix `/oh-my-claudecode:`.
 | `/oh-my-claudecode:ralph-init <task>` | Initialize PRD for structured task tracking |
 | `/oh-my-claudecode:ralph <task>` | Self-referential loop until task completion |
 | `/oh-my-claudecode:ultraqa <goal>` | Autonomous QA cycling workflow |
-| `/oh-my-claudecode:plan <description>` | Start planning session |
-| `/oh-my-claudecode:ralplan <description>` | Iterative planning with consensus |
+| `/oh-my-claudecode:plan <description>` | Start planning session (supports consensus structured deliberation) |
+| `/oh-my-claudecode:ralplan <description>` | Iterative planning with consensus structured deliberation (`--deliberate` for high-risk mode) |
 | `/oh-my-claudecode:review [plan-path]` | Review a plan with critic |
 | `/oh-my-claudecode:deepsearch <query>` | Thorough multi-strategy codebase search |
 | `/oh-my-claudecode:deepinit [path]` | Index codebase with hierarchical AGENTS.md files |
@@ -422,12 +446,12 @@ Just include these words anywhere in your prompt to activate enhanced modes:
 | Keyword | Effect |
 |---------|--------|
 | `ultrawork`, `ulw`, `uw` | Activates parallel agent orchestration |
-| ``, `eco`, `efficient`, `save-tokens`, `budget` | Token-efficient parallel execution |
+| `eco`, `efficient`, `save-tokens`, `budget` | Token-efficient parallel execution |
 | `autopilot`, `build me`, `I want a` | Full autonomous execution |
 | `ultrapilot`, `parallel build`, `swarm build` | Parallel autopilot (3-5x faster) |
 | `ralph`, `don't stop`, `must complete` | Persistence until verified complete |
 | `plan this`, `plan the` | Planning interview workflow |
-| `ralplan` | Iterative planning consensus |
+| `ralplan` | Iterative planning consensus with structured deliberation (`--deliberate` for high-risk mode) |
 | `search`, `find`, `locate` | Enhanced search mode |
 | `analyze`, `investigate`, `debug` | Deep analysis mode |
 | `sciomc` | Parallel research orchestration |
@@ -477,111 +501,20 @@ pipeline: analyze → fix → test this bug
 
 ---
 
-## MCP Path Boundary Rules
-
-The MCP tools (`ask_codex`, `ask_gemini`) enforce strict path boundaries for security. Both `prompt_file` and `output_file` are resolved relative to `working_directory`.
-
-### Default Behavior (Strict Mode)
-
-By default, both files must reside within the `working_directory`:
-
-| Parameter | Requirement |
-|-----------|-------------|
-| `prompt_file` | Must be within `working_directory` (after symlink resolution) |
-| `output_file` | Must be within `working_directory` (after symlink resolution) |
-| `working_directory` | Must be within the project worktree (unless bypassed) |
-
-### Environment Variable Overrides
-
-| Variable | Values | Description |
-|----------|--------|-------------|
-| `OMC_MCP_OUTPUT_PATH_POLICY` | `strict` (default), `redirect_output` | Controls output file path enforcement |
-| `OMC_MCP_OUTPUT_REDIRECT_DIR` | Path (default: `.omc/outputs`) | Directory for redirected outputs when policy is `redirect_output` |
-| `OMC_MCP_ALLOW_EXTERNAL_PROMPT` | `0` (default), `1` | Allow prompt files outside working directory |
-| `OMC_ALLOW_EXTERNAL_WORKDIR` | unset (default), `1` | Allow working_directory outside project worktree |
-| `OMC_DISCORD_WEBHOOK_URL` | URL | Discord webhook URL for notifications |
-| `OMC_DISCORD_NOTIFIER_BOT_TOKEN` | Token | Discord bot token for Bot API notifications |
-| `OMC_DISCORD_NOTIFIER_CHANNEL` | Channel ID | Discord channel ID for Bot API notifications |
-| `OMC_DISCORD_MENTION` | `<@uid>` or `<@&role_id>` | Mention to prepend to Discord messages |
-| `OMC_TELEGRAM_BOT_TOKEN` | Token | Telegram bot token for notifications |
-| `OMC_TELEGRAM_CHAT_ID` | Chat ID | Telegram chat ID for notifications |
-| `OMC_SLACK_WEBHOOK_URL` | URL | Slack incoming webhook URL for notifications |
-
-### Policy Descriptions
-
-**`OMC_MCP_OUTPUT_PATH_POLICY=strict` (Default)**
-- Output files must be within `working_directory`
-- Attempts to write outside the boundary fail with `E_PATH_OUTSIDE_WORKDIR_OUTPUT`
-- Most secure option - recommended for production
-
-**`OMC_MCP_OUTPUT_PATH_POLICY=redirect_output`**
-- Output files are automatically redirected to `OMC_MCP_OUTPUT_REDIRECT_DIR`
-- Only the filename is preserved; directory structure is flattened
-- Useful when you want to collect all outputs in a single location
-- Logs redirection at `[MCP Config]` level
-
-**`OMC_MCP_ALLOW_EXTERNAL_PROMPT=1`**
-- Allows reading prompt files from outside `working_directory`
-- **Security Warning**: Enables reading arbitrary files on the filesystem
-- Use only in trusted environments
-
-**`OMC_ALLOW_EXTERNAL_WORKDIR=1`**
-- Allows `working_directory` to be outside the project worktree
-- Bypasses the worktree boundary check
-- Use when running MCP tools against external projects
-
-### Error Tokens
-
-| Token | Meaning |
-|-------|---------|
-| `E_PATH_OUTSIDE_WORKDIR_PROMPT` | prompt_file is outside working_directory |
-| `E_PATH_OUTSIDE_WORKDIR_OUTPUT` | output_file is outside working_directory |
-| `E_PATH_RESOLUTION_FAILED` | Failed to resolve symlink or directory |
-| `E_WRITE_FAILED` | Failed to write output file (I/O error) |
-| `E_WORKDIR_INVALID` | working_directory does not exist or is inaccessible |
-
-### Example Valid/Invalid Patterns
-
-**Valid paths (working_directory: `/home/user/project`)**
-```
-prompt.txt                    -> /home/user/project/prompt.txt
-./prompts/task.md             -> /home/user/project/prompts/task.md
-../project/output.txt         -> /home/user/project/output.txt (resolves inside)
-.omc/outputs/response.md      -> /home/user/project/.omc/outputs/response.md
-```
-
-**Invalid paths (working_directory: `/home/user/project`)**
-```
-/etc/passwd                   -> Outside working directory (absolute)
-../../etc/shadow              -> Outside working directory (traverses too far)
-/tmp/output.txt               -> Outside working directory (different root)
-```
-
-### Troubleshooting Matrix
-
-| Symptom | Cause | Fix |
-|---------|-------|-----|
-| `E_PATH_OUTSIDE_WORKDIR_PROMPT` error | prompt_file outside working_directory | Move file to working directory or change working_directory to a common ancestor |
-| `E_PATH_OUTSIDE_WORKDIR_OUTPUT` error | output_file outside working_directory | Use relative path within working directory, or set `OMC_MCP_OUTPUT_PATH_POLICY=redirect_output` |
-| `E_PATH_RESOLUTION_FAILED` error | Symlink resolution failed or directory inaccessible | Ensure target directory exists and is accessible |
-| `E_WRITE_FAILED` error | I/O error (permissions, disk full) | Check file permissions and disk space |
-| `working_directory is outside the project worktree` | working_directory not within git worktree | Set `OMC_ALLOW_EXTERNAL_WORKDIR=1` or use a working directory inside the project |
-| Output file not where expected | `redirect_output` policy active | Check `OMC_MCP_OUTPUT_REDIRECT_DIR` (default: `.omc/outputs`) |
-
----
-
 ## Platform Support
 
 ### Operating Systems
 
 | Platform | Install Method | Hook Type |
 |----------|---------------|-----------|
-| **Windows** | `npm install -g` | Node.js (.mjs) |
+| **Windows** | WSL2 recommended (see note) | Node.js (.mjs) |
 | **macOS** | curl or npm | Bash (.sh) |
 | **Linux** | curl or npm | Bash (.sh) |
 
 > **Note**: Bash hooks are fully portable across macOS and Linux (no GNU-specific dependencies).
 
+> **Windows**: Native Windows (win32) support is experimental. OMC requires tmux, which is not available on native Windows. **WSL2 is strongly recommended** for Windows users. See the [WSL2 installation guide](https://learn.microsoft.com/en-us/windows/wsl/install). Native Windows issues may have limited support.
+
 > **Advanced**: Set `OMC_USE_NODE_HOOKS=1` to use Node.js hooks on macOS/Linux.
 
 ### Available Tools

package/hooks/hooks.json

@@ -7,12 +7,12 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/keyword-detector.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/keyword-detector.mjs\"",
             "timeout": 5
           },
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/skill-injector.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/skill-injector.mjs\"",
             "timeout": 3
           }
         ]
@@ -24,12 +24,12 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/session-start.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/session-start.mjs\"",
             "timeout": 5
           },
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/project-memory-session.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/project-memory-session.mjs\"",
             "timeout": 5
           }
         ]
@@ -39,7 +39,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/setup-init.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/setup-init.mjs\"",
             "timeout": 30
           }
         ]
@@ -49,7 +49,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/setup-maintenance.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/setup-maintenance.mjs\"",
             "timeout": 60
           }
         ]
@@ -61,17 +61,17 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/pre-tool-enforcer.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/pre-tool-enforcer.mjs\"",
             "timeout": 3
           }
         ]
       },
       {
-        "matcher": "TeamCreate|ExitPlanMode",
+        "matcher": "ExitPlanMode",
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/context-safety.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/context-safety.mjs\"",
             "timeout": 3
           }
         ]
@@ -83,7 +83,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/permission-handler.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/permission-handler.mjs\"",
             "timeout": 5
           }
         ]
@@ -95,12 +95,12 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/post-tool-verifier.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/post-tool-verifier.mjs\"",
             "timeout": 3
           },
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/project-memory-posttool.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/project-memory-posttool.mjs\"",
             "timeout": 3
           }
         ]
@@ -112,7 +112,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/post-tool-use-failure.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/post-tool-use-failure.mjs\"",
             "timeout": 3
           }
         ]
@@ -124,7 +124,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/subagent-tracker.mjs\" start",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/subagent-tracker.mjs\" start",
             "timeout": 3
           }
         ]
@@ -136,12 +136,12 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/subagent-tracker.mjs\" stop",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/subagent-tracker.mjs\" stop",
             "timeout": 5
           },
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/verify-deliverables.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/verify-deliverables.mjs\"",
             "timeout": 5
           }
         ]
@@ -153,12 +153,12 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/pre-compact.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/pre-compact.mjs\"",
             "timeout": 10
           },
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/project-memory-precompact.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/project-memory-precompact.mjs\"",
             "timeout": 5
           }
         ]
@@ -170,17 +170,17 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/context-guard-stop.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/context-guard-stop.mjs\"",
             "timeout": 5
           },
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/persistent-mode.cjs\"",
-            "timeout": 5
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/persistent-mode.cjs\"",
+            "timeout": 10
           },
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/code-simplifier.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/code-simplifier.mjs\"",
             "timeout": 5
           }
         ]
@@ -192,7 +192,7 @@
         "hooks": [
           {
             "type": "command",
-            "command": "sh \"${CLAUDE_PLUGIN_ROOT}/scripts/find-node.sh\" \"${CLAUDE_PLUGIN_ROOT}/scripts/session-end.mjs\"",
+            "command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/run.cjs\" \"${CLAUDE_PLUGIN_ROOT}/scripts/session-end.mjs\"",
             "timeout": 10
           }
         ]

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claudecode-omc",
-  "version": "4.4.4",
+  "version": "4.4.6",
   "description": "Multi-agent orchestration system for Claude Code - Inspired by oh-my-opencode",
   "type": "module",
   "main": "dist/index.js",
@@ -14,7 +14,6 @@
   "bin": {
     "oh-my-claudecode": "dist/cli/index.js",
     "omc": "dist/cli/index.js",
-    "omc-analytics": "dist/cli/analytics.js",
     "omc-cli": "dist/cli/index.js"
   },
   "files": [
@@ -22,9 +21,9 @@
     "agents",
     "bridge",
     "bridge/mcp-server.cjs",
-    "bridge/codex-server.cjs",
-    "bridge/gemini-server.cjs",
     "bridge/team-bridge.cjs",
+    "bridge/team-mcp.cjs",
+    "bridge/runtime-cli.cjs",
     "commands",
     "hooks",
     "scripts",
@@ -37,11 +36,13 @@
     "LICENSE"
   ],
   "scripts": {
-    "build": "tsc && node scripts/build-skill-bridge.mjs && node scripts/build-mcp-server.mjs && node scripts/build-codex-server.mjs && node scripts/build-gemini-server.mjs && node scripts/build-bridge-entry.mjs && npm run compose-docs",
+    "build": "tsc && node scripts/build-skill-bridge.mjs && node scripts/build-mcp-server.mjs && node scripts/build-bridge-entry.mjs && npm run compose-docs && npm run build:runtime-cli && npm run build:team-server",
     "build:bridge": "node scripts/build-skill-bridge.mjs",
-    "build:codex": "node scripts/build-codex-server.mjs",
-    "build:gemini": "node scripts/build-gemini-server.mjs",
+    "build:codex": "echo '[deprecated] Codex MCP server removed in v5.0 - use tmux team workers instead'",
+    "build:gemini": "echo '[deprecated] Gemini MCP server removed in v5.0 - use tmux team workers instead'",
     "build:bridge-entry": "node scripts/build-bridge-entry.mjs",
+    "build:runtime-cli": "node scripts/build-runtime-cli.mjs",
+    "build:team-server": "node scripts/build-team-server.mjs",
     "compose-docs": "node scripts/compose-docs.mjs",
     "dev": "tsc --watch",
     "start": "node dist/index.js",
@@ -61,7 +62,6 @@
     "@anthropic-ai/claude-agent-sdk": "^0.1.0",
     "@ast-grep/napi": "^0.31.0",
     "@modelcontextprotocol/sdk": "^1.26.0",
-    "@tokscale/core": "^1.0.25",
     "@types/better-sqlite3": "^7.6.13",
     "ajv": "^8.17.1",
     "better-sqlite3": "^12.6.2",