Skip to content
Back to search

multi-agent-shogun

Samurai-inspired multi-agent system for Claude Code. Orchestrate parallel AI tasks via tmux with shogun → karo → ashigaru hierarchy.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/yohey-w--multi-agent-shogun/badge.svg)](/listings/yohey-w--multi-agent-shogun)

HTML

<a href="/listings/yohey-w--multi-agent-shogun"><img src="/api/listings/yohey-w--multi-agent-shogun/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

**Requirements:** tmux, bash 4+, at least one of: [Claude Code](https://claude.ai/code) / Codex / Copilot / Kimi / OpenCode / Antigravity

git clone https://github.com/yohey-w/multi-agent-shogun
cd multi-agent-shogun
bash first_setup.sh                        # one-time setup: config, dependencies, MCP
source ~/.bashrc                           # reload PATH
claude --dangerously-skip-permissions      # first run only: OAuth + accept Bypass Permissions → /exit
bash shutsujin_departure.sh                # launch all agents

> For full install steps (incl. Windows) and the first-30-minutes walkthrough, see [🚀 Quick Start](#-quick-start) and the basic usage section below.

Type a command in the Shogun pane:

> "Build a REST API for user authentication"

Shogun delegates → Karo breaks it down → 7 Ashigaru execute in parallel. You watch the dashboard. That's it.

> **Want to go deeper?** The rest of this README covers architecture, configuration, memory design, and multi-CLI setup.

---

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 2,001 lines · 6 findings marked

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. <div align="center">
  2. # multi-agent-shogun
  3. **Command your AI army like a feudal warlord.**
  4. Run 10 AI coding agents in parallel — **Claude Code, OpenAI Codex, GitHub Copilot, Kimi Code, OpenCode, Cursor, Antigravity** — orchestrated through a samurai-inspired hierarchy with zero coordination overhead.
  5. **Talk Coding, not Vibe Coding. Speak to your phone, AI executes.**
  6. [![GitHub Stars](https://img.shields.io/github/stars/yohey-w/multi-agent-shogun?style=social)](https://github.com/yohey-w/multi-agent-shogun)
  7. [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
  8. [![v5.1.0 Karo Traffic Control](https://img.shields.io/badge/v5.1.0-Karo%20Traffic%20Control-ff6600?style=flat-square&logo=data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHdpZHRoPSIxNiIgaGVpZ2h0PSIxNiI+PHRleHQgeD0iMCIgeT0iMTIiIGZvbnQtc2l6ZT0iMTIiPuKalTwvdGV4dD48L3N2Zz4=)](https://github.com/yohey-w/multi-agent-shogun/releases/tag/v5.1.0)
  9. [![Shell](https://img.shields.io/badge/Shell%2FBash-100%25-green)]()
  10. [English](README.md) | [日本語](README_ja.md)
  11. </div>
  12. > **📖 Sequel (Aug 2026):** The author has since disbanded his own 10-agent army down to a single agent that carries his judgment. The successor project is [**kagemusha**](https://github.com/yohey-w/kagemusha) — it ships the forms of a *judgment loop* (corrections → principles → standing rules), not the agents. The story of the disbandment: [article (JP)](https://zenn.dev/shio_shoppaize/articles/kagemusha-shogun-disband). multi-agent-shogun itself continues to work as-is.
  13. <p align="center">
  14. <img src="images/screenshots/hero/latest-translucent-20260210-190453.png" alt="Latest translucent command session in the Shogun pane" width="940">
  15. </p>
  16. <p align="center">
  17. <img src="images/screenshots/hero/latest-translucent-20260208-084602.png" alt="Quick natural-language command in the Shogun pane" width="420">
  18. <img src="images/company-creed-all-panes.png" alt="Karo and Ashigaru panes reacting in parallel" width="520">
  19. </p>
  20. <p align="center"><i>One Karo (manager) coordinating 7 Ashigaru (workers) + 1 Gunshi (strategist) — real session, no mock data.</i></p>
  21. ---
  22. ## Quick Start
  23. **Requirements:** tmux, bash 4+, at least one of: [Claude Code](https://claude.ai/code) / Codex / Copilot / Kimi / OpenCode / Antigravity
  24. ```bash
  25. git clone https://github.com/yohey-w/multi-agent-shogun
  26. cd multi-agent-shogun
  27. bash first_setup.sh # one-time setup: config, dependencies, MCP
  28. source ~/.bashrc # reload PATH
  29. claude --dangerously-skip-permissions # first run only: OAuth + accept Bypass Permissions → /exit
  30. bash shutsujin_departure.sh # launch all agents
  31. ```
  32. > For full install steps (incl. Windows) and the first-30-minutes walkthrough, see [🚀 Quick Start](#-quick-start) and the basic usage section below.
  33. Type a command in the Shogun pane:
  34. > "Build a REST API for user authentication"
  35. Shogun delegates → Karo breaks it down → 7 Ashigaru execute in parallel.
  36. You watch the dashboard. That's it.
  37. > **Want to go deeper?** The rest of this README covers architecture, configuration, memory design, and multi-CLI setup.
  38. ---
  39. ## What is this?
  40. **multi-agent-shogun** is a system that runs multiple AI coding CLI instances simultaneously, orchestrating them like a feudal Japanese army. Supports **Claude Code**, **OpenAI Codex**, **GitHub Copilot**, **Kimi Code**, **OpenCode**, **Cursor**, and **Antigravity**.
  41. **Why use it?**
  42. - One command spawns 7 AI workers + 1 strategist executing in parallel
  43. - Zero wait time — give your next order while tasks run in the background
  44. - AI remembers your preferences across sessions (Memory MCP)
  45. - Real-time progress on a dashboard
  46. ```
  47. You (上様 / The Lord)
  48. │
  49. ▼ Give orders
  50. ┌─────────────┐
  51. │ SHOGUN │ ← Receives your command, delegates instantly
  52. └──────┬──────┘
  53. │ YAML + tmux
  54. ┌──────▼──────┐
  55. │ KARO │ ← Distributes tasks to workers
  56. └──────┬──────┘
  57. │
  58. ┌─┬─┬─┬─┴─┬─┬─┬─┬────────┐
  59. │1│2│3│4│5│6│7│ GUNSHI │ ← 7 workers + 1 strategist
  60. └─┴─┴─┴─┴─┴─┴─┴────────┘
  61. ASHIGARU 軍師
  62. ```
  63. ---
  64. ## Why Shogun?
  65. Most multi-agent frameworks burn API tokens on coordination. Shogun doesn't.
  66. | | Claude Code `Task` tool | Claude Code Agent Teams | LangGraph | CrewAI | **multi-agent-shogun** |
  67. |---|---|---|---|---|---|
  68. | **Architecture** | Subagents inside one process | Team lead + teammates (JSON mailbox) | Graph-based state machine | Role-based agents | Feudal hierarchy via tmux |
  69. | **Parallelism** | Sequential (one at a time) | Multiple independent sessions | Parallel nodes (v0.2+) | Limited | **8 independent agents** |
  70. | **Coordination cost** | API calls per Task | Token-heavy (each teammate = separate context) | API + infra (Postgres/Redis) | API + CrewAI platform | **Zero** (YAML + tmux) |
  71. | **Multi-CLI** | Claude Code only | Claude Code only | Any LLM API | Any LLM API | **7 CLIs** (Claude/Codex/Copilot/Kimi/OpenCode/Cursor/Antigravity) |
  72. | **Observability** | Claude logs only | tmux split-panes or in-process | LangSmith integration | OpenTelemetry | **Live tmux panes** + dashboard |
  73. | **Skill discovery** | None | None | None | None | **Bottom-up auto-proposal** |
  74. | **Setup** | Built into Claude Code | Built-in (experimental) | Heavy (infra required) | pip install | Shell scripts |
  75. ### What makes this different
  76. **Zero coordination overhead** — Agents talk through YAML files on disk. The only API calls are for actual work, not orchestration. Run 8 agents and pay only for 8 agents' work.
  77. **Full transparency** — Every agent runs in a visible tmux pane. Every instruction, report, and decision is a plain YAML file you can read, diff, and version-control. No black boxes.
  78. **Battle-tested hierarchy** — The Shogun → Karo → Ashigaru chain of command prevents conflicts by design: clear ownership, dedicated files per agent, event-driven communication, no polling.
  79. ---
  80. ## Why CLI (Not API)?
  81. Most AI coding tools charge per token. Running 8 Opus-grade agents through the API costs **$100+/hour**. CLI subscriptions flip this:
  82. | | API (Per-Token) | CLI (Flat-Rate) |
  83. |---|---|---|
  84. | **8 agents × Opus** | ~$100+/hour | ~$200/month |
  85. | **Cost predictability** | Unpredictable spikes | Fixed monthly bill |
  86. | **Usage anxiety** | Every token counts | Unlimited |
  87. | **Experimentation budget** | Constrained | Deploy freely |
  88. **"Use AI recklessly"** — With flat-rate CLI subscriptions, deploy 8 agents without hesitation. The cost is the same whether they work 1 hour or 24 hours. No more choosing between "good enough" and "thorough" — just run more agents.
  89. ### Multi-CLI Support
  90. Shogun isn't locked to one vendor. The system supports 7 CLI tools, each with unique strengths:
  91. | CLI | Key Strength | Default Model |
  92. |-----|-------------|---------------|
  93. | **Claude Code** | Battle-tested tmux integration, Memory MCP, dedicated file tools (Read/Write/Edit/Glob/Grep) | Claude Sonnet 4.6 |
  94. | **OpenAI Codex** | Sandbox execution, JSONL structured output, `codex exec` headless mode, **per-model `--model` flag** | gpt-5.3-codex / **gpt-5.3-codex-spark** |
  95. | **GitHub Copilot** | Built-in GitHub MCP, 4 specialized agents (Explore/Task/Plan/Code-review), `/delegate` to coding agent | Claude Sonnet 4.6 |
  96. | **Kimi Code** | Free tier available, strong multilingual support | Kimi k2 |
  97. | **OpenCode** | Shared `AGENTS.md` instructions, agent-specific definitions via `--agent`, `/new` context reset, restart-only model changes, deterministic interactive TUI launch, provider-qualified `--model` routing | provider/model |
  98. | **Cursor** | Auto-loads `CLAUDE.md`/`AGENTS.md`/`.cursor/rules/`, built-in web search, `inbox-write` skill via `.cursor/skills/`, `/model` live switching, `--yolo` auto-run | Varies |
  99. | **Antigravity CLI** | Google Antigravity CLI integration via `agy`, host-managed auth, YOLO-style launch, `gemini`/`agy` legacy aliases | host default / last-used |
  100. OpenCode sessions load the agent-specific `.opencode/agents/<agent_id>.md` definition via `--agent` and keep automation resets on `/new`; model changes require a relaunch. Automation uses the repository-provided `config/opencode-tui.json` via `OPENCODE_TUI_CONFIG`, which disables `app_exit` and pins `session_interrupt`/`input_clear` to known bindings. Role boundaries are embedded in the generated agent frontmatter: Shogun can read `queue/reports/*` for oversight but cannot write them, Karo is limited to coordination files plus report aggregation, Ashigaru only touch their own task/report pair, and Gunshi reads ashigaru reports but only writes `gunshi_report.yaml`.
  101. Antigravity sessions launch with `agy --dangerously-skip-permissions`. Shogun treats `type: antigravity`, `type: agy`, and legacy `type: gemini` as Antigravity. Authentication and default model selection stay in the host user's Antigravity CLI setup; `settings.yaml` may optionally pass a concrete `model`, but `auto` uses the host default or last-used model.
  102. A unified instruction build system generates CLI-specific instruction files from shared templates:
  103. ```
  104. instructions/
  105. ├── common/ # Shared rules (all CLIs)
  106. ├── cli_specific/ # CLI-specific tool descriptions
  107. │ ├── claude_tools.md # Claude Code tools & features
  108. │ ├── copilot_tools.md # GitHub Copilot CLI tools & features
  109. │ ├── opencode_tools.md # OpenCode tools, agent frontmatter, and permission model
  110. │ └── cursor_tools.md # Cursor Agent tools, skills, and session rules
  111. └── roles/ # Role definitions (shogun, karo, ashigaru)
  112. ↓ build
  113. CLAUDE.md / AGENTS.md / .github/copilot-instructions.md / .opencode/agents/*.md / .cursor/rules/*.md
  114. ← Generated per CLI
  115. ```
  116. One source of truth, zero sync drift. Change a rule once, all CLIs get it.
  117. ---
  118. ## Bottom-Up Skill Discovery
  119. This is the feature no other framework has.
  120. As Ashigaru execute tasks, they **automatically identify reusable patterns** and propose them as skill candidates. The Karo aggregates these proposals in `dashboard.md`, and you — the Lord — decide what gets promoted to a permanent skill.
  121. ```
  122. Ashigaru finishes a task
  123. ↓
  124. Notices: "I've done this pattern 3 times across different projects"
  125. ↓
  126. Reports in YAML: skill_candidate:
  127. found: true
  128. name: "api-endpoint-scaffold"
  129. reason: "Same REST scaffold pattern used in 3 projects"
  130. ↓
  131. Appears in dashboard.md → You approve → Skill created in .claude/commands/
  132. ↓
  133. Any agent can now invoke /api-endpoint-scaffold
  134. ```
  135. Skills grow organically from real work — not from a predefined template library. Your skill set becomes a reflection of **your** workflow.
  136. ---
  137. ## Quick Start
  138. ### Windows (WSL2)
  139. <table>
  140. <tr>
  141. <td width="60">
  142. **Step 1**
  143. </td>
  144. <td>
  145. 📥 **Download the repository**
  146. [Download ZIP](https://github.com/yohey-w/multi-agent-shogun/archive/refs/heads/main.zip) and extract to `C:\tools\multi-agent-shogun`
  147. *Or use git:* `git clone https://github.com/yohey-w/multi-agent-shogun.git C:\tools\multi-agent-shogun`
  148. </td>
  149. </tr>
  150. <tr>
  151. <td>
  152. **Step 2**
  153. </td>
  154. <td>
  155. 🖱️ **Run `install.bat`**
  156. Right-click → "Run as Administrator" (if WSL2 is not installed). Sets up WSL2 + Ubuntu automatically.
  157. </td>
  158. </tr>
  159. <tr>
  160. <td>
  161. **Step 3**
  162. </td>
  163. <td>
  164. 🐧 **Open Ubuntu and run** (first time only)
  165. ```bash
  166. cd /mnt/c/tools/multi-agent-shogun
  167. ./first_setup.sh
  168. ```
  169. </td>
  170. </tr>
  171. <tr>
  172. <td>
  173. **Step 4**
  174. </td>
  175. <td>
  176. ✅ **Deploy!**
  177. ```bash
  178. ./shutsujin_departure.sh
  179. ```
  180. </td>
  181. </tr>
  182. </table>
  183. #### First-time only: Authentication
  184. After `first_setup.sh`, run these commands once to authenticate:
  185. ```bash
  186. # 1. Apply PATH changes
  187. source ~/.bashrc
  188. # 2. OAuth login + Bypass Permissions approval (one command)
  189. claude --dangerously-skip-permissions
  190. # → Browser opens → Log in with Anthropic account → Return to CLI
  191. # → "Bypass Permissions" prompt appears → Select "Yes, I accept" (↓ to option 2, Enter)
  192. # → Type /exit to quit
  193. ```
  194. This saves credentials to `~/.claude/` — you won't need to do it again.
  195. #### Daily startup
  196. Open an **Ubuntu terminal** (WSL) and run:
  197. ```bash
  198. cd /mnt/c/tools/multi-agent-shogun
  199. ./shutsujin_departure.sh
  200. ```
  201. ### 📱 Mobile Access — Dedicated Android App (Recommended)
  202. <p align="center">
  203. <img src="android/screenshots/01_shogun_terminal.png" alt="Shogun Terminal" width="200">
  204. <img src="android/screenshots/02_agents_grid.png" alt="Agents Grid" width="200">
  205. <img src="android/screenshots/03_dashboard.png" alt="Dashboard" width="200">
  206. </p>
  207. Monitor and command 10 AI agents from your phone with the dedicated Android companion app.
  208. | Feature | Description |
  209. |---------|-------------|
  210. | **Shogun Terminal** | SSH terminal + voice input + special key bar (C-c, C-b, Tab, etc.) |
  211. | **Agents Grid** | 9-pane simultaneous monitoring. Tap to expand fullscreen + send commands |
  212. | **Dashboard** | Renders dashboard.md with full table text selection/copy |
  213. | **Rate Limit** | Tap the FAB on the Agents tab to check Claude Max 5h/7d usage with progress bars |
  214. | **Voice Input** | Japanese continuous recognition via Google Speech API — higher accuracy than phone keyboard voice |
  215. | **Screenshot Share** | Share images via Android share menu → SFTP transfer to server |
  216. > **Note:** Android only for now. No iOS version — the developer doesn't own an iPhone. If there's demand, please open an [Issue](https://github.com/yohey-w/multi-agent-shogun/issues). PRs welcome!
  217. #### Setup
  218. **Prerequisites:**
  219. - Shogun system running on WSL2 (or Linux server)
  220. - SSH server started (`sudo service ssh start`)
  221. - Phone and server on same network (LAN or [Tailscale](https://tailscale.com/))
  222. **Steps:**
  223. 1. **Install APK**
  224. 1. Download [`android/release/multi-agent-shogun.apk`](android/release/multi-agent-shogun.apk) on your phone (open the file on GitHub → "Download raw file")
  225. 2. Tap the download notification → "Install"
  226. 3. If "Unknown sources" warning appears → "Settings" → enable "Allow from this source" for your browser → go back → "Install"
  227. 4. Done → "Open"
  228. 2. **Configure SSH** (Settings tab)
  229. | Field | Example | Description |
  230. |-------|---------|-------------|
  231. | SSH Host | `100.xxx.xxx.xxx` | Server IP (e.g., Tailscale IP) |
  232. | SSH Port | `22` | Usually 22 |
  233. | SSH User | `your_username` | SSH login username |
  234. | SSH Key Path | `/data/data/.../id_ed25519` | Private key path on phone (*1) |
  235. | SSH Password | `****` | Use if no key available |
  236. | Project Path | `/mnt/c/tools/multi-agent-shogun` | Server-side project directory |
  237. | Shogun Session | `shogun` | tmux session name for Shogun |
  238. | Agent Session | `multiagent` | tmux session name for agents |
  239. *1 Transfer your private key to the phone, or use password authentication
  240. 3. **Save → Switch to Shogun tab** → auto-connects
  241. **Using Tailscale (connect from anywhere):**
  242. ```bash
  243. # Server-side (WSL2)
  244. curl -fsSL https://tailscale.com/install.sh | sh
  245. sudo tailscaled &
  246. sudo tailscale up --authkey tskey-auth-XXXXXXXXXXXX
  247. sudo service ssh start
  248. ```
  249. Install the Tailscale app on your phone, log in with the same account, and use the displayed Tailscale IP as the SSH Host in the app.
  250. **With ntfy notifications:**
  251. See [ntfy setup section](#-8-phone-notifications-ntfy) for push notifications from Karo on task completion.
  252. <details>
  253. <summary>📟 <b>Termux Method (without the Android app)</b> (click to expand)</summary>
  254. SSH via Termux also works. More limited than the dedicated app, but requires no APK sideloading.
  255. **Requirements (all free):**
  256. | Name | In a nutshell | Role |
  257. |------|--------------|------|
  258. | [Tailscale](https://tailscale.com/) | A road to your home from anywhere | Connect to your home PC from anywhere |
  259. | SSH | The feet that walk that road | Log into your home PC through Tailscale |
  260. | [Termux](https://termux.dev/) | A black screen on your phone | Required to use SSH — just install it |
  261. **Setup:**
  262. 1. Install Tailscale on both WSL and your phone
  263. 2. In WSL (auth key method — browser not needed):
  264. ```bash
  265. curl -fsSL https://tailscale.com/install.sh | sh
  266. sudo tailscaled &
  267. sudo tailscale up --authkey tskey-auth-XXXXXXXXXXXX
  268. sudo service ssh start
  269. ```
  270. 3. In Termux on your phone:
  271. ```sh
  272. pkg update && pkg install openssh
  273. ssh youruser@your-tailscale-ip
  274. css # Connect to Shogun
  275. ```
  276. 4. Open a new Termux window (+ button) for workers:
  277. ```sh
  278. ssh youruser@your-tailscale-ip
  279. csm # See all 9 panes
  280. ```
  281. **Disconnect:** Just swipe the Termux window closed. tmux sessions survive — agents keep working.
  282. </details>
  283. ---
  284. <details>
  285. <summary>🐧 <b>Linux / macOS</b> (click to expand)</summary>
  286. ### First-time setup
  287. ```bash
  288. # 1. Clone
  289. git clone https://github.com/yohey-w/multi-agent-shogun.git ~/multi-agent-shogun
  290. cd ~/multi-agent-shogun
  291. # 2. Make scripts executable
  292. chmod +x *.sh
  293. # 3. Run first-time setup
  294. ./first_setup.sh
  295. ```
  296. ### Daily startup
  297. ```bash
  298. cd ~/multi-agent-shogun
  299. ./shutsujin_departure.sh
  300. ```
  301. </details>
  302. ---
  303. <details>
  304. <summary>❓ <b>What is WSL2? Why is it needed?</b> (click to expand)</summary>
  305. ### About WSL2
  306. **WSL2 (Windows Subsystem for Linux)** lets you run Linux inside Windows. This system uses `tmux` (a Linux tool) to manage multiple AI agents, so WSL2 is required on Windows.
  307. ### If you don't have WSL2 yet
  308. No problem! Running `install.bat` will:
  309. 1. Check if WSL2 is installed (auto-install if not)
  310. 2. Check if Ubuntu is installed (auto-install if not)
  311. 3. Guide you through next steps (running `first_setup.sh`)
  312. **Quick install command** (run PowerShell as Administrator):
  313. ```powershell
  314. wsl --install
  315. ```
  316. Then restart your computer and run `install.bat` again.
  317. </details>
  318. ---
  319. <details>
  320. <summary>📋 <b>Script Reference</b> (click to expand)</summary>
  321. | Script | Purpose | When to run |
  322. |--------|---------|-------------|
  323. | `install.bat` | Windows: WSL2 + Ubuntu setup | First time only |
  324. | `first_setup.sh` | Install tmux, Node.js, Claude Code CLI + Memory MCP config | First time only |
  325. | `shutsujin_departure.sh` | Create tmux sessions + launch the configured CLI for each agent + load instructions + start ntfy listener | Daily |
  326. | `scripts/switch_cli.sh` | Live switch agent CLI/model (settings.yaml → /exit → relaunch) | As needed |
  327. ### What `install.bat` does automatically:
  328. - ✅ Checks if WSL2 is installed (guides you if not)
  329. - ✅ Checks if Ubuntu is installed (guides you if not)
  330. - ✅ Shows next steps (how to run `first_setup.sh`)
  331. ### What `shutsujin_departure.sh` does:
  332. - ✅ Creates tmux sessions (shogun + multiagent)
  333. - ✅ Launches each agent with the CLI configured in `config/settings.yaml` (Claude/Codex/Copilot/Kimi/OpenCode)
  334. - ✅ Auto-loads instruction files or generated agent definitions for each CLI
  335. - ✅ Resets queue files for a fresh state
  336. - ✅ Starts ntfy listener for phone notifications (if configured)
  337. **After running, all agents are ready to receive commands!**
  338. </details>
  339. ---
  340. <details>
  341. <summary>🔧 <b>Manual Requirements</b> (click to expand)</summary>
  342. If you prefer to install dependencies manually:
  343. | Requirement | Installation | Notes |
  344. |-------------|-------------|-------|
  345. | WSL2 + Ubuntu | `wsl --install` in PowerShell | Windows only |
  346. | Set Ubuntu as default | `wsl --set-default Ubuntu` | Required for scripts to work |
  347. | tmux | `sudo apt install tmux` | Terminal multiplexer |
  348. | Node.js v20+ | `nvm install 20` | Required for MCP servers |
  349. | Claude Code CLI | `curl -fsSL https://claude.ai/install.sh \| bash` | Official Anthropic CLI (native version recommended; npm version deprecated) |
  350. | OpenAI Codex CLI | Install from the official OpenAI Codex distribution | Required only for agents with `type: codex` |
  351. | GitHub Copilot CLI | Install and authenticate GitHub Copilot CLI | Required only for agents with `type: copilot` |
  352. | Kimi Code CLI | Install and authenticate Kimi Code | Required only for agents with `type: kimi` |
  353. | OpenCode CLI | `npm install -g opencode-ai` | Required only for agents with `type: opencode`; provider API keys must be available in the agent shell |
  354. | Cursor CLI | See [Cursor CLI docs](https://cursor.com/docs/cli/overview) — use `cursor-agent` or `agent` command | Required only for agents with `type: cursor` |
  355. | Antigravity CLI | Install and authenticate Google Antigravity CLI (`agy`) | Required only for agents with `type: antigravity`, `type: agy`, or legacy `type: gemini` |
  356. </details>
  357. ---
  358. ### After Setup
  359. Whichever option you chose, **10 AI agents** are automatically launched:
  360. | Agent | Role | Count |
  361. |-------|------|-------|
  362. | 🏯 Shogun | Supreme commander — receives your orders | 1 |
  363. | 📋 Karo | Manager — distributes tasks, quality checks | 1 |
  364. | ⚔️ Ashigaru | Workers — execute implementation tasks in parallel | 7 |
  365. | 🧠 Gunshi | Strategist — handles analysis, evaluation, and design | 1 |
  366. Two tmux sessions are created:
  367. - `shogun` — connect here to give commands
  368. - `multiagent` — Karo, Ashigaru, and Gunshi running in the background
  369. ---
  370. ## How It Works
  371. ### Step 1: Connect to the Shogun
  372. After running `shutsujin_departure.sh`, all agents automatically load their instructions and are ready.
  373. Open a new terminal and connect:
  374. ```bash
  375. tmux attach-session -t shogun
  376. ```
  377. ### Step 2: Give your first order
  378. The Shogun is already initialized — just give a command:
  379. ```
  380. Research the top 5 JavaScript frameworks and create a comparison table
  381. ```
  382. The Shogun will:
  383. 1. Write the task to a YAML file
  384. 2. Notify the Karo (manager)
  385. 3. Return control to you immediately — no waiting!
  386. Meanwhile, the Karo distributes tasks to Ashigaru workers for parallel execution.
  387. ### Step 3: Check progress
  388. Open `dashboard.md` in your editor for a real-time status view:
  389. ```markdown
  390. ## In Progress
  391. | Worker | Task | Status |
  392. |--------|------|--------|
  393. | Ashigaru 1 | Research React | Running |
  394. | Ashigaru 2 | Research Vue | Running |
  395. | Ashigaru 3 | Research Angular | Completed |
  396. ```
  397. ### Project-Unit Operation (Equivalent to Visual Studio "Solution")
  398. Once set up, the Shogun system can handle **multiple projects under the same Shogun**, switching between them as needed. The unit equivalent to a Visual Studio "solution" is `projects/{name}.yaml` + `context/{name}.md`.
  399. #### 1. Running your first project
  400. ```bash
  401. # (1) Connect to the Shogun (after shutsujin_departure.sh completes)
  402. tmux attach-session -t shogun
  403. # (2) Just give the Shogun your command — the project starts automatically
  404. # → Shogun writes cmd to queue/shogun_to_karo.yaml and notifies Karo
  405. # → Karo distributes to Ashigaru for parallel execution
  406. # → Results aggregate in dashboard.md
  407. ```
  408. No explicit "create a project" command is needed. The Shogun attaches a `project:` field to the cmd when relevant, and related files are automatically separated.
  409. #### 2. Explicitly registering a project (optional, for long-term work)
  410. For ongoing projects, you can place metadata in `projects/{name}.yaml`:
  411. ```yaml
  412. # projects/example.yaml
  413. id: example
  414. name: "Sample Project"
  415. working_directory: /path/to/repo
  416. north_star: "The ultimate goal for this project"
  417. notes: |
  418. Project-specific notes, stakeholders, special rules
  419. ```
  420. The Shogun and Karo reference this file and inject project context when issuing cmds.
  421. Detailed project knowledge (requirements, design, past feedback) lives in `context/{name}.md`. When the Shogun issues a cmd related to the project, it automatically references this file.
  422. #### 3. Customizing the agent formation
  423. The agent formation (which CLI each agent uses) lives in `config/settings.yaml`:
  424. ```yaml
  425. cli:
  426. agents:
  427. ashigaru1:
  428. type: codex # codex / claude / copilot / kimi / opencode / antigravity
  429. model: gpt-5.5
  430. ashigaru2:
  431. type: claude
  432. model: claude-sonnet-4-6
  433. # Same for ashigaru3-7, gunshi, karo
  434. ```
  435. OpenCode uses provider-qualified model IDs:
  436. ```yaml
  437. cli:
  438. agents:
  439. ashigaru3:
  440. type: opencode
  441. model: openrouter/openai/gpt-4o-mini
  442. variant: high # optional provider-specific reasoning variant
  443. ```
  444. OpenRouter setup has two separate pieces:
  445. 1. **Model routing** goes in `config/settings.yaml` as shown above (`type: opencode`, `model: openrouter/...`).
  446. 2. **Provider authentication** is configured in OpenCode, not in `settings.yaml`. Run OpenCode once as the same OS user that will launch Shogun, then use `/connect` → `OpenRouter` and paste the API key. OpenCode stores provider credentials in its own user data under that OS user (for example under `~/.local/share/opencode/`; the exact file/database is OpenCode-internal). For headless deployments that use environment-based provider credentials, make sure the shell that runs `shutsujin_departure.sh` has `OPENROUTER_API_KEY` loaded.
  447. Do not put API keys in `config/settings.yaml`, `config/opencode-tui.json`, or `.opencode/agents/*.md`. Those files only describe routing, tmux-safe keybindings, and generated agent definitions.
  448. When OpenCode is selected, `lib/cli_adapter.sh` launches it with `--agent <agent_id>` and the repository-pinned `OPENCODE_TUI_CONFIG=config/opencode-tui.json`. The TUI command does not accept `--variant`; if `variant:` is configured, `scripts/build_instructions.sh` and `scripts/switch_cli.sh` synchronize `model:` / `variant:` into a git-ignored `.opencode/agents/<agent_id>-runtime.md`, which OpenCode loads via `--agent <agent_id>-runtime`.
  449. To switch on the fly, use `scripts/switch_cli.sh`:
  450. ```bash
  451. bash scripts/switch_cli.sh ashigaru3 --type claude --model claude-sonnet-4-6
  452. bash scripts/switch_cli.sh ashigaru3 --type opencode --model openrouter/openai/gpt-4o-mini
  453. bash scripts/switch_cli.sh ashigaru3 --type opencode --model openrouter/minimax/minimax-m2.5 --variant xhigh
  454. ```
  455. #### 4. Switching or closing a project
  456. There is no explicit "close project" command. **Issuing the next project's cmd automatically switches context.**
  457. - Pause temporarily: do nothing. Old cmds remain in `queue/` as history, and the Shogun restores state when resumed
  458. - Fully retire: delete `projects/{name}.yaml`, or add an `archived: true` flag
  459. - Run in parallel: use the `project:` field in cmds to keep concurrent projects distinct
  460. #### 5. Carrying experience and settings between projects
  461. What carries forward to future projects:
  462. | What carries forward | Stored in | Referenced when |
  463. |----------------------|-----------|-----------------|
  464. | Lord's preferences and lessons | Memory MCP (persistent) | All agents at Session Start |
  465. | Project-specific knowledge | `context/{name}.md` | When running the project's cmds |
  466. | Past cmd history | `queue/shogun_to_karo.yaml` | When the Shogun needs it |
  467. | Custom skills | `~/.claude/skills/`, `skills/` | When matching triggers fire |
  468. | Agent formation | `config/settings.yaml` | At shutsujin startup |
  469. **Memory MCP** is the heart of "experience." When you tell the Shogun "don't do X next time" or "remember Y," the Shogun records it in Memory MCP, and all future projects see it.
  470. ### Detailed flow
  471. ```
  472. You: "Research the top 5 MCP servers and create a comparison table"
  473. ```
  474. The Shogun writes the task to `queue/shogun_to_karo.yaml` and wakes the Karo. Control returns to you immediately.
  475. The Karo breaks the task into subtasks:
  476. | Worker | Assignment |
  477. |--------|-----------|
  478. | Ashigaru 1 | Research Notion MCP |
  479. | Ashigaru 2 | Research GitHub MCP |
  480. | Ashigaru 3 | Research Playwright MCP |
  481. | Ashigaru 4 | Research Memory MCP |
  482. | Ashigaru 5 | Research Sequential Thinking MCP |
  483. All 5 Ashigaru research simultaneously. You can watch them work in real time:
  484. <p align="center">
  485. <img src="images/company-creed-all-panes.png" alt="Ashigaru agents working in parallel across tmux panes" width="900">
  486. </p>
  487. Results appear in `dashboard.md` as they complete.
  488. ---
  489. ## Key Features
  490. ### ⚡ 1. Parallel Execution
  491. One command spawns up to 8 parallel tasks:
  492. ```
  493. You: "Research 5 MCP servers"
  494. → 5 Ashigaru start researching simultaneously
  495. → Results in minutes, not hours
  496. ```
  497. ### 🔄 2. Non-Blocking Workflow
  498. The Shogun delegates instantly and returns control to you:
  499. ```
  500. You: Command → Shogun: Delegates → You: Give next command immediately
  501. ↓
  502. Workers: Execute in background
  503. ↓
  504. Dashboard: Shows results
  505. ```
  506. No waiting for long tasks to finish.
  507. ### 🧠 3. Cross-Session Memory (Memory MCP)
  508. Your AI remembers your preferences:
  509. ```
  510. Session 1: Tell it "I prefer simple approaches"
  511. → Saved to Memory MCP
  512. Session 2: AI loads memory on startup
  513. → Stops suggesting complex solutions
  514. ```
  515. ### 📡 4. Event-Driven Communication (Zero Polling)
  516. Agents talk to each other by writing YAML files — like passing notes. **No polling loops, no wasted API calls.**
  517. ```
  518. Karo wants to wake Ashigaru 3:
  519. Step 1: Write the message Step 2: Wake the agent up
  520. ┌──────────────────────┐ ┌──────────────────────────┐
  521. │ inbox_write.sh │ │ inbox_watcher.sh │
  522. │ │ │ │
  523. │ Writes full message │ file │ Detects file change │
  524. │ to ashigaru3.yaml │──change──▶│ (inotifywait, not poll) │
  525. │ with flock (no race) │ │ │
  526. └──────────────────────┘ │ Wakes agent via: │
  527. │ 1. Self-watch (skip) │
  528. │ 2. tmux send-keys │
  529. │ (short nudge only) │
  530. └──────────────────────────┘
  531. Step 3: Agent reads its own inbox
  532. ┌──────────────────────────────────┐
  533. │ Ashigaru 3 reads ashigaru3.yaml │
  534. │ → Finds unread messages │
  535. │ → Processes them │
  536. │ → Marks as read │
  537. └──────────────────────────────────┘
  538. ```
  539. **How the wake-up works:**
  540. | Priority | Method | What happens | When used |
  541. |----------|--------|-------------|-----------|
  542. | 1st | **Self-Watch** | Agent watches its own inbox file — wakes itself, no nudge needed | Agent has its own `inotifywait` running |
  543. | 2nd | **Stop Hook** | Claude Code agents check inbox at turn end via `.claude/settings.json` Stop hook | Claude Code agents only |
  544. | 3rd | **tmux send-keys** | Sends short nudge via `tmux send-keys` (text and Enter sent separately for Codex CLI compatibility) | Fallback — disabled in ASW Phase 2+ |
  545. **Agent Self-Watch (ASW) Phases** — Controls how aggressively the system uses `tmux send-keys` nudges:
  546. | ASW Phase | Nudge behavior | Delivery method | When to use |
  547. |-----------|---------------|-----------------|-------------|
  548. | **Phase 1** | Normal nudges enabled | self-watch + send-keys | Initial setup, mixed CLI environments |
  549. | **Phase 2** | **Busy → suppressed, Idle → nudge** | busy: stop hook delivers at turn end. idle: nudge (unavoidable) | Claude Code agents with stop hook (recommended) |
  550. | **Phase 3** | `FINAL_ESCALATION_ONLY` | send-keys only as last-resort recovery | Fully stable environments |
  551. Phase 2 uses the idle flag file (`/tmp/shogun_idle_{agent}`) to distinguish busy vs idle agents. The Stop hook creates/removes this flag at turn boundaries. This eliminates nudge interruptions during active work while still waking idle agents.
  552. > **Why can't nudges be fully eliminated?** Claude Code's Stop hook only fires at turn end. An idle agent (sitting at the prompt) has no turn ending, so there's no hook to trigger inbox checks. A future `Notification` hook with `idle_prompt` blocking support or a periodic timer hook could solve this.
  553. Configure in `config/settings.yaml`:
  554. ```yaml
  555. asw_phase: 2 # Recommended for Claude Code setups
  556. ```
  557. Or set the default directly in `scripts/inbox_watcher.sh` (`ASW_PHASE` variable). Restart inbox_watcher processes after changing.
  558. **3-Phase Escalation (v3.2)** — If agent doesn't respond:
  559. | Phase | Timing | Action |
  560. |-------|--------|--------|
  561. | Phase 1 | 0-2 min | Standard nudge (`inbox3` text + Enter) — *skipped for busy agents in ASW Phase 2+* |
  562. | Phase 2 | 2-4 min | Copilot/Kimi: Escape×2 + single Ctrl-C + nudge. Claude/Codex/OpenCode: plain nudge fallback |
  563. | Phase 3 | 4+ min | Send CLI-specific context reset: Claude/Copilot/Kimi use `/clear`, Codex/OpenCode use `/new` (max once per 5 min) |
  564. **Key design choices:**
  565. - **Message content is never sent through tmux** — only a short "you have mail" nudge. The agent reads its own file. This eliminates character corruption and transmission hangs.
  566. - **Zero CPU while idle** — `inotifywait` blocks on a kernel event (not a poll loop). CPU usage is 0% between messages.
  567. - **Guaranteed delivery** — If the file write succeeded, the message is there. No lost messages, no retries needed.
  568. ### 📊 5. Agent Status Check
  569. See which agents are busy or idle — instantly, from one command:
  570. ```bash
  571. # Project mode: full status with task/inbox info
  572. bash scripts/agent_status.sh
  573. # Standalone mode: works with any tmux session
  574. bash scripts/agent_status.sh --session mysession --lang en
  575. ```
  576. **Project mode output:**
  577. ```
  578. Agent CLI Pane Task ID Status Inbox
  579. ---------- ------- --------- ------------------------------------------ ---------- -----
  580. karo claude 待機中 --- --- 0
  581. ashigaru1 codex 稼働中 subtask_042a_research assigned 0
  582. ashigaru2 codex 待機中 subtask_042b_review done 0
  583. gunshi claude 稼働中 subtask_042c_analysis assigned 0
  584. ```
  585. **Standalone mode output** (no project config needed):
  586. ```
  587. Pane State Agent ID
  588. ------------------------------ ---------- ----------
  589. multiagent:agents.0 IDLE karo
  590. multiagent:agents.1 BUSY ashigaru1
  591. multiagent:agents.8 BUSY gunshi
  592. ```
  593. Detection works for **Claude Code**, **Codex CLI**, and **OpenCode** by checking CLI-specific prompt/spinner patterns near the bottom of each tmux pane. The detection logic lives in `lib/agent_status.sh` — source it in your own scripts:
  594. ```bash
  595. source lib/agent_status.sh
  596. agent_is_busy_check "multiagent:agents.3" && echo "busy" || echo "idle"
  597. ```
  598. ### 📸 6. Screenshot Integration
  599. VSCode's Claude Code extension lets you paste screenshots to explain issues. This CLI system provides the same capability:
  600. ```yaml
  601. # Set your screenshot folder in config/settings.yaml
  602. screenshot:
  603. path: "/mnt/c/Users/YourName/Pictures/Screenshots"
  604. ```
  605. ```
  606. # Just tell the Shogun:
  607. You: "Check the latest screenshot"
  608. You: "Look at the last 2 screenshots"
  609. → AI instantly reads and analyzes your screen captures
  610. ```
  611. **Windows tip:** Press `Win + Shift + S` to take screenshots. Set the save path in `settings.yaml` for seamless integration.
  612. Use cases:
  613. - Explain UI bugs visually
  614. - Show error messages
  615. - Compare before/after states
  616. ### 📁 7. Context Management (4-Layer Architecture)
  617. Efficient knowledge sharing through a four-layer context system:
  618. | Layer | Location | Purpose |
  619. |-------|----------|---------|
  620. | Layer 1: Memory MCP | `memory/shogun_memory.jsonl` | Cross-project, cross-session long-term memory |
  621. | Layer 2: Project | `config/projects.yaml`, `projects/<id>.yaml`, `context/{project}.md` | Project-specific information and technical knowledge |
  622. | Layer 3: YAML Queue | `queue/shogun_to_karo.yaml`, `queue/tasks/`, `queue/reports/` | Task management — source of truth for instructions and reports |
  623. | Layer 4: Session | CLAUDE.md, instructions/*.md | Working context (wiped by `/clear`) |
  624. #### Persistent Agent Memory (`memory/MEMORY.md`)
  625. Shogun reads `memory/MEMORY.md` at every session start. It contains Lord's preferences, lessons learned, and cross-session knowledge — written by Shogun, read by Shogun.
  626. ```
  627. ┌─────────────────────────────────────────────────────────────┐
  628. │ Git Repositories │
  629. │ │
  630. │ ┌─────────────────────┐ ┌──────────────────────────┐ │
  631. │ │ multi-agent-shogun │ │ shogun-private │ │
  632. │ │ (public OSS) │ │ (your private repo) │ │
  633. │ │ │ │ │ │
  634. │ │ scripts/ │ │ projects/client.yaml ←──┐ │ │
  635. │ │ instructions/ │ │ context/my-notes.md ←──┤ │ │
  636. │ │ lib/ │ │ queue/shogun_to_karo.yaml │ │ │
  637. │ │ memory/ │ │ memory/MEMORY.md ←──┘ │ │
  638. │ │ ├─ MEMORY.md.sample│ │ config/settings.yaml │ │
  639. │ │ └─ MEMORY.md ─────┼───┼── same file, tracked here │ │
  640. │ │ (gitignored) │ │ │ │
  641. │ └─────────────────────┘ └──────────────────────────┘ │
  642. │ ↑ anyone can fork ↑ your data, your repo │
  643. └─────────────────────────────────────────────────────────────┘
  644. ```
  645. **How it works:** `memory/MEMORY.md` lives in the same working directory as the OSS repo, but is excluded from the OSS `.gitignore` (whitelist-based). You track it in a separate private repo using a bare git repo technique:
  646. ```bash
  647. # One-time setup (already done by first_setup.sh)
  648. git init --bare ~/.shogun-private.git
  649. alias privategit='git --git-dir=$HOME/.shogun-private.git --work-tree=/path/to/multi-agent-shogun'
  650. privategit remote add origin https://github.com/YOU/shogun-private.git
  651. # Daily use
  652. privategit add -f memory/MEMORY.md projects/my-client.yaml
  653. privategit commit -m "update memory"
  654. privategit push
  655. ```
  656. The OSS `.gitignore` uses a **whitelist approach** (default: exclude everything, then explicitly allow OSS files). So private files like `memory/MEMORY.md` are automatically excluded without needing explicit `gitignore` entries — just don't add them to the whitelist.
  657. This design enables:
  658. - Any Ashigaru can work on any project
  659. - Context persists across agent switches
  660. - Clear separation of concerns
  661. - Knowledge survives across sessions
  662. #### /clear Protocol (Cost Optimization)
  663. As agents work, their session context (Layer 4) grows, increasing API costs. `/clear` wipes session memory and resets costs. Layers 1–3 persist as files, so nothing is lost.
  664. Recovery cost after `/clear`: **~6,800 tokens** (42% improved from v1 — CLAUDE.md YAML conversion + English-only instructions reduced token cost by 70%)
  665. 1. CLAUDE.md (auto-loaded) → recognizes itself as part of the Shogun System
  666. 2. `tmux display-message -t "$TMUX_PANE" -p '#{@agent_id}'` → identifies its own number
  667. 3. Memory MCP read → restores the Lord's preferences (~700 tokens)
  668. 4. Task YAML read → picks up the next assignment (~800 tokens)
  669. The key insight: designing **what not to load** is what drives cost savings.
  670. #### Universal Context Template
  671. All projects use the same 7-section template:
  672. | Section | Purpose |
  673. |---------|---------|
  674. | What | Project overview |
  675. | Why | Goals and success criteria |
  676. | Who | Stakeholders and responsibilities |
  677. | Constraints | Deadlines, budgets, limitations |
  678. | Current State | Progress, next actions, blockers |
  679. | Decisions | Decisions made and their rationale |
  680. | Notes | Free-form observations and ideas |
  681. This unified format enables:
  682. - Quick onboarding for any agent
  683. - Consistent information management across all projects
  684. - Easy handoff between Ashigaru workers
  685. ### 📱 8. Phone Notifications (ntfy)
  686. Two-way communication between your phone and the Shogun — no SSH, no Tailscale, no server needed.
  687. | Direction | How it works |
  688. |-----------|-------------|
  689. | **Phone → Shogun** | Send a message from the ntfy app → `ntfy_listener.sh` receives it via streaming → auto-ACK reply (`📱受信: {your message}`) sent back to your phone → Shogun processes automatically |
  690. | **Karo → Phone (direct)** | When Karo updates `dashboard.md`, it sends push notifications directly via `scripts/ntfy.sh` — **Shogun is bypassed** (Shogun is for human interaction, not progress reporting) |
  691. ```
  692. 📱 You (from bed) 🏯 Shogun
  693. │ │
  694. │ "Research React 19" │
  695. ├─────────────────────────►│
  696. │ (ntfy message) │ → Delegates to Karo → Ashigaru work
  697. │ │
  698. │ "✅ cmd_042 complete" │
  699. │◄─────────────────────────┤
  700. │ (push notification) │
  701. ```
  702. **Setup:**
  703. 1. Add `ntfy_topic: "shogun-yourname"` to `config/settings.yaml`
  704. 2. Install the [ntfy app](https://ntfy.sh) on your phone and subscribe to the same topic
  705. 3. `shutsujin_departure.sh` automatically starts the listener — no extra steps
  706. **Notification examples:**
  707. | Event | Notification |
  708. |-------|-------------|
  709. | Command completed | `✅ cmd_042 complete — 5/5 subtasks done` |
  710. | Task failed | `❌ subtask_042c failed — API rate limit` |
  711. | Action required | `🚨 Action needed: approve skill candidate` |
  712. | Streak update | `🔥 3-day streak! 12/12 tasks today` |
  713. Free, no account required, no server to maintain. Uses [ntfy.sh](https://ntfy.sh) — an open-source push notification service.
  714. > **⚠️ Security:** Your topic name is your password. Anyone who knows it can read your notifications and send messages to your Shogun. Choose a hard-to-guess name and **never share it publicly** (e.g., in screenshots, blog posts, or GitHub commits).
  715. **Verify it works:**
  716. ```bash
  717. # Send a test notification to your phone
  718. bash scripts/ntfy.sh "Test notification from Shogun 🏯"
  719. ```
  720. If your phone receives the notification, you're all set. If not, check:
  721. - `config/settings.yaml` has `ntfy_topic` set (not empty, no extra quotes)
  722. - The ntfy app on your phone is subscribed to **the exact same topic name**
  723. - Your phone has internet access and ntfy notifications are enabled
  724. **Sending commands from your phone:**
  725. 1. Open the ntfy app on your phone
  726. 2. Tap your subscribed topic
  727. 3. Type a message (e.g., `Research React 19 best practices`) and send
  728. 4. `ntfy_listener.sh` receives it, writes to `queue/ntfy_inbox.yaml`, and wakes the Shogun
  729. 5. The Shogun reads the message and processes it through the normal Karo → Ashigaru pipeline
  730. Any text you send becomes a command. Write it like you'd talk to the Shogun — no special syntax needed.
  731. **Manual listener start** (if not using `shutsujin_departure.sh`):
  732. ```bash
  733. # Start the listener in the background
  734. nohup bash scripts/ntfy_listener.sh &>/dev/null &
  735. # Check if it's running
  736. pgrep -f ntfy_listener.sh
  737. # View listener logs (stderr output)
  738. bash scripts/ntfy_listener.sh # Run in foreground to see logs
  739. ```
  740. The listener automatically reconnects if the connection drops. `shutsujin_departure.sh` starts it automatically on deployment — you only need manual start if you skipped the deployment script.
  741. **Troubleshooting:**
  742. | Problem | Fix |
  743. |---------|-----|
  744. | No notifications on phone | Check topic name matches exactly in `settings.yaml` and ntfy app |
  745. | Listener not starting | Run `bash scripts/ntfy_listener.sh` in foreground to see errors |
  746. | Phone → Shogun not working | Verify listener is running: `pgrep -f ntfy_listener.sh` |
  747. | Messages not reaching Shogun | Check `queue/ntfy_inbox.yaml` — if message is there, Shogun may be busy |
  748. | "ntfy_topic not configured" error | Add `ntfy_topic: "your-topic"` to `config/settings.yaml` |
  749. | Duplicate notifications | Normal on reconnect — Shogun deduplicates by message ID |
  750. | Changed topic name but no notifications | The listener must be restarted: `pkill -f ntfy_listener.sh && nohup bash scripts/ntfy_listener.sh &>/dev/null &` |
  751. **Real-world notification screenshots:**
  752. <p align="center">
  753. <img src="images/screenshots/masked/ntfy_saytask_rename.jpg" alt="Bidirectional phone communication" width="300">
  754. &nbsp;&nbsp;
  755. <img src="images/screenshots/masked/ntfy_cmd043_progress.jpg" alt="Progress notification" width="300">
  756. </p>
  757. <p align="center"><i>Left: Bidirectional phone ↔ Shogun communication · Right: Real-time progress report from Ashigaru</i></p>
  758. <p align="center">
  759. <img src="images/screenshots/masked/ntfy_bloom_oc_test.jpg" alt="Command completion notification" width="300">
  760. &nbsp;&nbsp;
  761. <img src="images/screenshots/masked/ntfy_persona_eval_complete.jpg" alt="8-agent parallel completion" width="300">
  762. </p>
  763. <p align="center"><i>Left: Command completion notification · Right: All 8 Ashigaru completing in parallel</i></p>
  764. > *Note: Topic names shown in screenshots are examples. Use your own unique topic name.*
  765. #### SayTask Notifications
  766. Behavioral psychology-driven motivation through your notification feed:
  767. - **Streak tracking**: Consecutive completion days counted in `saytask/streaks.yaml` — maintaining streaks leverages loss aversion to sustain momentum
  768. - **Eat the Frog** 🐸: The hardest task of the day is marked as the "Frog." Completing it triggers a special celebration notification
  769. - **Daily progress**: `12/12 tasks today` — visual completion feedback reinforces the Arbeitslust effect (joy of work-in-progress)
  770. ### 🖼️ 9. Pane Border Task Display
  771. Each tmux pane shows the agent's current task directly on its border:
  772. ```
  773. ┌ ashigaru1 Sonnet+T VF requirements ──┬ ashigaru3 Opus+T API research ──────┐
  774. │ │ │
  775. │ Working on SayTask requirements │ Researching REST API patterns │
  776. │ │ │
  777. ├ ashigaru2 Sonnet ───────────────────┼ ashigaru4 Spark DB schema design ───┤
  778. │ │ │
  779. │ (idle — waiting for assignment) │ Designing database schema │
  780. │ │ │
  781. └──────────────────────────────────────┴─────────────────────────────────────┘
  782. ```
  783. - **Working**: `ashigaru1 Sonnet+T VF requirements` — agent name, model (with Thinking indicator), and task summary
  784. - **Idle**: `ashigaru2 Sonnet` — model name only, no task
  785. - **Display names**: Sonnet, Opus, Haiku, Codex, Spark — `+T` suffix = Extended Thinking enabled
  786. - Updated automatically by the Karo when assigning or completing tasks
  787. - Glance at all 9 panes to instantly know who's doing what
  788. ### 🔊 10. Shout Mode (Battle Cries)
  789. When an Ashigaru completes a task, it shouts a personalized battle cry in the tmux pane — a visual reminder that your army is working hard.
  790. ```
  791. ┌ ashigaru1 (Sonnet) ──────────┬ ashigaru2 (Sonnet) ──────────┐
  792. │ │ │
  793. │ ⚔️ 足軽1号、先陣切った! │ 🔥 足軽2号、二番槍の意地! │
  794. │ 八刃一志! │ 八刃一志! │
  795. │ ❯ │ ❯ │
  796. └───────────────────────────────┴───────────────────────────────┘
  797. ```
  798. **How it works:**
  799. The Karo writes an `echo_message` field in each task YAML. After completing all work (report + inbox notification), the Ashigaru runs `echo` as its **final action**. The message stays visible above the `❯` prompt.
  800. ```yaml
  801. # In the task YAML (written by Karo)
  802. task:
  803. task_id: subtask_001
  804. description: "Create comparison table"
  805. echo_message: "🔥 足軽1号、先陣を切って参る!八刃一志!"
  806. ```
  807. **Shout mode is the default.** To disable (saves API tokens on the echo call):
  808. ```bash
  809. ./shutsujin_departure.sh --silent # No battle cries
  810. ./shutsujin_departure.sh # Default: shout mode (battle cries enabled)
  811. ```
  812. Silent mode sets `DISPLAY_MODE=silent` as a tmux environment variable. The Karo checks this when writing task YAMLs and omits the `echo_message` field.
  813. ---
  814. ## 🗣️ SayTask — Task Management for People Who Hate Task Management
  815. ### What is SayTask?
  816. **Task management for people who hate task management. Just speak to your phone.**
  817. **Talk Coding, not Vibe Coding.** Speak your tasks, AI organizes them. No typing, no opening apps, no friction.
  818. - **Target audience**: People who installed Todoist but stopped opening it after 3 days
  819. - Your enemy isn't other apps — it's doing nothing. The competition is inaction, not another productivity tool
  820. - Zero UI. Zero typing. Zero app-opening. Just talk
  821. > *"Your enemy isn't other apps — it's doing nothing."*
  822. ### How it Works
  823. 1. Install the [ntfy app](https://ntfy.sh) (free, no account needed)
  824. 2. Speak to your phone: *"dentist tomorrow"*, *"invoice due Friday"*
  825. 3. AI auto-organizes → morning notification: *"here's your day"*
  826. ```
  827. 🗣️ "Buy milk, dentist tomorrow, invoice due Friday"
  828. │
  829. ▼
  830. ┌──────────────────┐
  831. │ ntfy → Shogun │ AI auto-categorize, parse dates, set priorities
  832. └────────┬─────────┘
  833. │
  834. ▼
  835. ┌──────────────────┐
  836. │ tasks.yaml │ Structured storage (local, never leaves your machine)
  837. └────────┬─────────┘
  838. │
  839. ▼
  840. 📱 Morning notification:
  841. "Today: 🐸 Invoice due · 🦷 Dentist 3pm · 🛒 Buy milk"
  842. ```
  843. ### Before / After
  844. | Before (v1) | After (v2) |
  845. |:-----------:|:----------:|
  846. | ![Task list v1](images/screenshots/masked/ntfy_tasklist_v1_before.jpg) | ![Task list v2](images/screenshots/masked/ntfy_tasklist_v2_aligned.jpg) |
  847. | Raw task dump | Clean, organized daily summary |
  848. > *Note: Topic names shown in screenshots are examples. Use your own unique topic name.*
  849. ### Use Cases
  850. - 🛏️ **In bed**: *"Gotta submit the report tomorrow"* — captured before you forget, no fumbling for a notebook
  851. - 🚗 **While driving**: *"Don't forget the estimate for client A"* — hands-free, eyes on the road
  852. - 💻 **Mid-work**: *"Oh, need to buy milk"* — dump it instantly and stay in flow
  853. - 🌅 **Wake up**: Today's tasks already waiting in your notifications — no app to open, no inbox to check
  854. - 🐸 **Eat the Frog**: AI picks your hardest task each morning — ignore it or conquer it first
  855. ### FAQ
  856. **Q: How is this different from other task apps?**
  857. A: You never open an app. Just speak. Zero friction. Most task apps fail because people stop opening them. SayTask removes that step entirely.
  858. **Q: Can I use SayTask without the full Shogun system?**
  859. A: SayTask is a feature of Shogun. Shogun also works as a standalone multi-agent development platform — you get both capabilities in one system.
  860. **Q: What's the Frog 🐸?**
  861. A: Every morning, AI picks your hardest task — the one you'd rather avoid. Tackle it first (the "Eat the Frog" method) or ignore it. Your call.
  862. **Q: Is it free?**
  863. A: Everything is free and open-source. ntfy is free too. No account, no server, no subscription.
  864. **Q: Where is my data stored?**
  865. A: Local YAML files on your machine. Nothing is sent to the cloud. Your tasks never leave your device.
  866. **Q: What if I say something vague like "that thing for work"?**
  867. A: AI does its best to categorize and schedule it. You can always refine later — but the point is capturing the thought before it disappears.
  868. ### SayTask vs cmd Pipeline
  869. Shogun has two complementary task systems:
  870. | Capability | SayTask (Voice Layer) | cmd Pipeline (AI Execution) |
  871. |---|:-:|:-:|
  872. | Voice input → task creation | ✅ | — |
  873. | Morning notification digest | ✅ | — |
  874. | Eat the Frog 🐸 selection | ✅ | — |
  875. | Streak tracking | ✅ | ✅ |
  876. | AI-executed tasks (multi-step) | — | ✅ |
  877. | 8-agent parallel execution | — | ✅ |
  878. SayTask handles personal productivity (capture → schedule → remind). The cmd pipeline handles complex work (research, code, multi-step tasks). Both share streak tracking — completing either type of task counts toward your daily streak.
  879. ---
  880. ## Model Settings
  881. | Agent | Default Model | Thinking | Role |
  882. |-------|--------------|----------|------|
  883. | Shogun | Opus | **Enabled (high)** | Strategic advisor to the Lord. Use `--shogun-no-thinking` for relay-only mode |
  884. | Karo | Sonnet | Enabled | Task distribution, simple QC, dashboard management |
  885. | Gunshi | Opus | Enabled | Deep analysis, design review, architecture evaluation |
  886. | Ashigaru 1–7 | Sonnet 4.6 | Enabled | Implementation: code, research, file operations |
  887. **Thinking control**: Set `thinking: true/false` per agent in `config/settings.yaml`. When `thinking: false`, the agent starts with `MAX_THINKING_TOKENS=0` to disable Extended Thinking. Pane borders show `+T` suffix when Thinking is enabled (e.g., `Sonnet+T`, `Opus+T`).
  888. **Live model switching**: Use `/shogun-model-switch` to change any agent's CLI type, model, or Thinking setting without restarting the entire system. See the Skills section for details.
  889. The system routes work by **cognitive complexity** at two levels: **Agent routing** (Ashigaru for L1–L3, Gunshi for L4–L6) and **Model routing within Ashigaru** via `capability_tiers` (see Dynamic Model Routing below).
  890. ### Bloom's Taxonomy → Agent Routing
  891. Tasks are classified using Bloom's Taxonomy and routed to the appropriate **agent**, not model:
  892. | Level | Category | Description | Routed To |
  893. |-------|----------|-------------|-----------|
  894. | L1 | Remember | Recall facts, copy, list | **Ashigaru** |
  895. | L2 | Understand | Explain, summarize, paraphrase | **Ashigaru** |
  896. | L3 | Apply | Execute procedures, implement known patterns | **Ashigaru** |
  897. | L4 | Analyze | Compare, investigate, deconstruct | **Gunshi** |
  898. | L5 | Evaluate | Judge, critique, recommend | **Gunshi** |
  899. | L6 | Create | Design, build, synthesize new solutions | **Gunshi** |
  900. The Karo assigns each subtask a Bloom level and routes it to the appropriate agent. L1–L3 tasks go to Ashigaru for parallel execution; L4–L6 tasks go to the Gunshi for deeper analysis. Simple L4 tasks (e.g., small code review) may still go to Ashigaru when the Karo judges it appropriate.
  901. ### Task Dependencies (blockedBy)
  902. Tasks can declare dependencies on other tasks using `blockedBy`:
  903. ```yaml
  904. # queue/tasks/ashigaru2.yaml
  905. task:
  906. task_id: subtask_010b
  907. blockedBy: ["subtask_010a"] # Waits for ashigaru1's task to complete
  908. description: "Integrate the API client built by subtask_010a"
  909. ```
  910. When a blocking task completes, the Karo automatically unblocks dependent tasks and assigns them to available Ashigaru. This prevents idle waiting and enables efficient pipelining of dependent work.
  911. ### Dynamic Model Routing (capability_tiers)
  912. Beyond agent-level routing, you can configure **model-level routing within the Ashigaru tier**. Define a `capability_tiers` table in `config/settings.yaml` mapping each model to its maximum Bloom level:
  913. ```yaml
  914. capability_tiers:
  915. gpt-5.3-codex-spark:
  916. max_bloom: 3 # L1–L3 only: fast, high-volume tasks
  917. cost_group: chatgpt_pro
  918. gpt-5.3-codex:
  919. max_bloom: 4 # L1–L4: + analysis and debugging
  920. cost_group: chatgpt_pro
  921. claude-sonnet-4-6:
  922. max_bloom: 5 # L1–L5: + design evaluation
  923. cost_group: claude_max
  924. claude-opus-4-6:
  925. max_bloom: 6 # L1–L6: + novel architecture, strategy
  926. cost_group: claude_max
  927. ```
  928. The `cost_group` field links each model to your subscription plan, enabling the system to avoid routing tasks to models your plan doesn't cover.
  929. Two built-in skills help you configure this:
  930. | Skill | Purpose |
  931. |-------|---------|
  932. | `/shogun-model-list` | Reference table: all models × subscriptions × Bloom max |
  933. | `/shogun-bloom-config` | Interactive: answer 2 questions → get ready-to-paste YAML |
  934. Run `/shogun-bloom-config` after setup to generate your optimal `capability_tiers` configuration.
  935. ---
  936. ## Philosophy
  937. > "Don't execute tasks mindlessly. Always keep 'fastest × best output' in mind."
  938. The Shogun System is built on five core principles:
  939. | Principle | Description |
  940. |-----------|-------------|
  941. | **Autonomous Formation** | Design task formations based on complexity, not templates |
  942. | **Parallelization** | Use subagents to prevent single-point bottlenecks |
  943. | **Research First** | Search for evidence before making decisions |
  944. | **Continuous Learning** | Don't rely solely on model knowledge cutoffs |
  945. | **Triangulation** | Multi-perspective research with integrated authorization |
  946. These principles are documented in detail: **[docs/philosophy.md](docs/philosophy.md)**
  947. ---
  948. ## Design Philosophy
  949. ### Why a hierarchy (Shogun → Karo → Ashigaru)?
  950. 1. **Instant response**: The Shogun delegates immediately, returning control to you
  951. 2. **Parallel execution**: The Karo distributes to multiple Ashigaru simultaneously
  952. 3. **Single responsibility**: Each role is clearly separated — no confusion
  953. 4. **Scalability**: Adding more Ashigaru doesn't break the structure
  954. 5. **Fault isolation**: One Ashigaru failing doesn't affect the others
  955. 6. **Unified reporting**: Only the Shogun communicates with you, keeping information organized
  956. ### Why Mailbox System?
  957. Why use files instead of direct messaging between agents?
  958. | Problem with direct messaging | How mailbox solves it |
  959. |-------------------------------|----------------------|
  960. | Agent crashes → message lost | YAML files survive restarts |
  961. | Polling wastes API calls | `inotifywait` is event-driven (zero CPU while idle) |
  962. | Agents interrupt each other | Each agent has its own inbox file — no cross-talk |
  963. | Hard to debug | Open any `.yaml` file to see exact message history |
  964. | Concurrent writes corrupt data | `flock` (exclusive lock) serializes writes automatically |
  965. | Delivery failures (character corruption, hangs) | Message content stays in files — only a short "you have mail" nudge is sent through tmux |
  966. ### Agent Identification (@agent_id)
  967. Each pane has a `@agent_id` tmux user option (e.g., `karo`, `ashigaru1`). While `pane_index` can shift when panes are rearranged, `@agent_id` is set at startup by `shutsujin_departure.sh` and never changes.
  968. Agent self-identification:
  969. ```bash
  970. tmux display-message -t "$TMUX_PANE" -p '#{@agent_id}'
  971. ```
  972. The `-t "$TMUX_PANE"` is required. Omitting it returns the active pane's value (whichever pane you're focused on), causing misidentification.
  973. Model names are stored as `@model_name` and current task summaries as `@current_task` — both displayed in the `pane-border-format`. Even if Claude Code overwrites the pane title, these user options persist.
  974. ### Why only the Karo updates dashboard.md
  975. 1. **Single writer**: Prevents conflicts by limiting updates to one agent
  976. 2. **Information aggregation**: The Karo receives all Ashigaru reports, so it has the full picture
  977. 3. **Consistency**: All updates pass through a single quality gate
  978. 4. **No interruptions**: If the Shogun updated it, it could interrupt the Lord's input
  979. ---
  980. ## Skills
  981. No skills are included out of the box. Skills emerge organically during operation — you approve candidates from `dashboard.md` as they're discovered.
  982. Invoke skills with `/skill-name`. Just tell the Shogun: "run /skill-name".
  983. ### Included Skills (committed to repo)
  984. Skills ship with the repository in `skills/`. They are domain-agnostic utilities useful for any user:
  985. | Skill | Description |
  986. |-------|-------------|
  987. | `/skill-creator` | Template and guide for creating new skills |
  988. | `/shogun-agent-status` | Show busy/idle status of all agents with task and inbox info |
  989. | `/shogun-model-list` | Reference table: all CLI tools × models × subscriptions × Bloom max level |
  990. | `/shogun-bloom-config` | Interactive configurator: answer 2 questions about your subscriptions → get ready-to-paste `capability_tiers` YAML |
  991. | `/shogun-model-switch` | Live CLI/model switching: settings.yaml update → `/exit` → relaunch with correct flags. Supports Thinking ON/OFF control |
  992. | `/shogun-readme-sync` | Keep README.md and README_ja.md in sync |
  993. These help you configure and operate the system. Personal workflow skills grow organically through the bottom-up discovery process.
  994. ### Skill Philosophy
  995. **1. Personal skills are not committed to the repo**
  996. Skills in `.claude/commands/` are excluded from version control by design:
  997. - Every user's workflow is different
  998. - Rather than imposing generic skills, each user grows their own skill set
  999. **2. How skills are discovered**
  1000. ```
  1001. Ashigaru notices a pattern during work
  1002. ↓
  1003. Appears in dashboard.md under "Skill Candidates"
  1004. ↓
  1005. You (the Lord) review the proposal
  1006. ↓
  1007. If approved, instruct the Karo to create the skill
  1008. ```
  1009. Skills are user-driven. Automatic creation would lead to unmanageable bloat — only keep what you find genuinely useful.
  1010. ---
  1011. ## MCP Setup Guide
  1012. MCP (Model Context Protocol) servers extend Claude's capabilities. Here's how to set them up:
  1013. ### What is MCP?
  1014. MCP servers give Claude access to external tools:
  1015. - **Notion MCP** → Read and write Notion pages
  1016. - **GitHub MCP** → Create PRs, manage issues
  1017. - **Memory MCP** → Persist memory across sessions
  1018. ### Installing MCP Servers
  1019. Add MCP servers with these commands:
  1020. ```bash
  1021. # 1. Notion - Connect to your Notion workspace
  1022. claude mcp add notion -e NOTION_TOKEN=your_token_here -- npx -y @notionhq/notion-mcp-server
  1023. # 2. Playwright - Browser automation
  1024. claude mcp add playwright -- npx @playwright/mcp@latest
  1025. # Note: Run `npx playwright install chromium` first
  1026. # 3. GitHub - Repository operations
  1027. claude mcp add github -e GITHUB_PERSONAL_ACCESS_TOKEN=your_pat_here -- npx -y @modelcontextprotocol/server-github
  1028. # 4. Sequential Thinking - Step-by-step reasoning for complex problems
  1029. claude mcp add sequential-thinking -- npx -y @modelcontextprotocol/server-sequential-thinking
  1030. # 5. Memory - Cross-session long-term memory (recommended!)
  1031. # ✅ Auto-configured by first_setup.sh
  1032. # To reconfigure manually:
  1033. claude mcp add memory -e MEMORY_FILE_PATH="$PWD/memory/shogun_memory.jsonl" -- npx -y @modelcontextprotocol/server-memory
  1034. ```
  1035. ### Verify installation
  1036. ```bash
  1037. claude mcp list
  1038. ```
  1039. All servers should show "Connected" status.
  1040. ---
  1041. ## Real-World Use Cases
  1042. This system manages **all white-collar tasks**, not just code. Projects can live anywhere on your filesystem.
  1043. ### Example 1: Research sprint
  1044. ```
  1045. You: "Research the top 5 AI coding assistants and compare them"
  1046. What happens:
  1047. 1. Shogun delegates to Karo
  1048. 2. Karo assigns:
  1049. - Ashigaru 1: Research GitHub Copilot
  1050. - Ashigaru 2: Research Cursor
  1051. - Ashigaru 3: Research Claude Code
  1052. - Ashigaru 4: Research Codeium
  1053. - Ashigaru 5: Research Amazon CodeWhisperer
  1054. 3. All 5 research simultaneously
  1055. 4. Results compiled in dashboard.md
  1056. ```
  1057. ### Example 2: PoC preparation
  1058. ```
  1059. You: "Prepare a PoC for the project on this Notion page: [URL]"
  1060. What happens:
  1061. 1. Karo fetches Notion content via MCP
  1062. 2. Ashigaru 2: Lists items to verify
  1063. 3. Ashigaru 3: Investigates technical feasibility
  1064. 4. Ashigaru 4: Drafts a PoC plan
  1065. 5. All results compiled in dashboard.md — meeting prep done
  1066. ```
  1067. ---
  1068. ## Configuration
  1069. ### Language
  1070. ```yaml
  1071. # config/settings.yaml
  1072. language: ja # Samurai Japanese only
  1073. language: en # Samurai Japanese + English translation
  1074. ```
  1075. ### Screenshot integration
  1076. ```yaml
  1077. # config/settings.yaml
  1078. screenshot:
  1079. path: "/mnt/c/Users/YourName/Pictures/Screenshots"
  1080. ```
  1081. Tell the Shogun "check the latest screenshot" and it reads your screen captures for visual context. (`Win+Shift+S` on Windows.)
  1082. ### ntfy (Phone Notifications)
  1083. ```yaml
  1084. # config/settings.yaml
  1085. ntfy_topic: "shogun-yourname"
  1086. ```
  1087. Subscribe to the same topic in the [ntfy app](https://ntfy.sh) on your phone. The listener starts automatically with `shutsujin_departure.sh`.
  1088. #### ntfy Authentication (Self-Hosted Servers)
  1089. The public ntfy.sh instance requires **no authentication** — the setup above is all you need.
  1090. If you run a self-hosted ntfy server with access control enabled, configure authentication:
  1091. ```bash
  1092. # 1. Copy the sample config
  1093. cp config/ntfy_auth.env.sample config/ntfy_auth.env
  1094. # 2. Edit with your credentials (choose one method)
  1095. ```
  1096. | Method | Config | When to use |
  1097. |--------|--------|-------------|
  1098. | **Bearer Token** (recommended) | `NTFY_TOKEN=tk_your_token_here` | Self-hosted ntfy with token auth (`ntfy token add <user>`) |
  1099. | **Basic Auth** | `NTFY_USER=username` + `NTFY_PASS=password` | Self-hosted ntfy with user/password |
  1100. | **None** (default) | Leave file empty or don't create it | Public ntfy.sh — no auth needed |
  1101. Priority: Token > Basic > None. If neither is set, no auth headers are sent (backward compatible).
  1102. `config/ntfy_auth.env` is excluded from git. See `config/ntfy_auth.env.sample` for details.
  1103. ---
  1104. ## Advanced
  1105. <details>
  1106. <summary><b>Script Architecture</b> (click to expand)</summary>
  1107. ```
  1108. ┌─────────────────────────────────────────────────────────────────────┐
  1109. │ First-Time Setup (run once) │
  1110. ├─────────────────────────────────────────────────────────────────────┤
  1111. │ │
  1112. │ install.bat (Windows) │
  1113. │ │ │
  1114. │ ├── Check/guide WSL2 installation │
  1115. │ └── Check/guide Ubuntu installation │
  1116. │ │
  1117. │ first_setup.sh (run manually in Ubuntu/WSL) │
  1118. │ │ │
  1119. │ ├── Check/install tmux │
  1120. │ ├── Check/install Node.js v20+ (via nvm) │
  1121. │ ├── Check/install Claude Code CLI (native version) │
  1122. │ │ ※ Proposes migration if npm version detected │
  1123. │ └── Configure Memory MCP server │
  1124. │ │
  1125. ├─────────────────────────────────────────────────────────────────────┤
  1126. │ Daily Startup (run every day) │
  1127. ├─────────────────────────────────────────────────────────────────────┤
  1128. │ │
  1129. │ shutsujin_departure.sh │
  1130. │ │ │
  1131. │ ├──▶ Create tmux sessions │
  1132. │ │ • "shogun" session (1 pane) │
  1133. │ │ • "multiagent" session (9 panes, 3x3 grid) │
  1134. │ │ │
  1135. │ ├──▶ Reset queue files and dashboard │
  1136. │ │ │
  1137. │ └──▶ Launch the configured CLI for each agent │
  1138. │ │
  1139. └─────────────────────────────────────────────────────────────────────┘
  1140. ```
  1141. </details>
  1142. <details>
  1143. <summary><b>shutsujin_departure.sh Options</b> (click to expand)</summary>
  1144. ```bash
  1145. # Default: Full startup (tmux sessions + configured CLI launch)
  1146. ./shutsujin_departure.sh
  1147. # Session setup only (no CLI launch)
  1148. ./shutsujin_departure.sh -s
  1149. ./shutsujin_departure.sh --setup-only
  1150. # Clean task queues (preserves command history)
  1151. ./shutsujin_departure.sh -c
  1152. ./shutsujin_departure.sh --clean
  1153. # Battle formation: All Ashigaru on Opus (max capability, higher cost)
  1154. ./shutsujin_departure.sh -k
  1155. ./shutsujin_departure.sh --kessen
  1156. # Silent mode: Disable battle cries (saves API tokens on echo calls)
  1157. ./shutsujin_departure.sh -S
  1158. ./shutsujin_departure.sh --silent
  1159. # Full startup + open Windows Terminal tabs
  1160. ./shutsujin_departure.sh -t
  1161. ./shutsujin_departure.sh --terminal
  1162. # Shogun relay-only mode: Disable Shogun's thinking (cost savings)
  1163. ./shutsujin_departure.sh --shogun-no-thinking
  1164. # Show help
  1165. ./shutsujin_departure.sh -h
  1166. ./shutsujin_departure.sh --help
  1167. ```
  1168. </details>
  1169. <details>
  1170. <summary><b>Common Workflows</b> (click to expand)</summary>
  1171. **Normal daily use:**
  1172. ```bash
  1173. ./shutsujin_departure.sh # Launch everything
  1174. tmux attach-session -t shogun # Connect and give commands
  1175. ```
  1176. **Debug mode (manual control):**
  1177. ```bash
  1178. ./shutsujin_departure.sh -s # Create sessions only
  1179. # Manually launch Claude Code on specific agents
  1180. tmux send-keys -t shogun:0 'claude --dangerously-skip-permissions' Enter
  1181. tmux send-keys -t multiagent:0.0 'claude --dangerously-skip-permissions' Enter
  1182. ```
  1183. **Restart after crash:**
  1184. ```bash
  1185. # Kill existing sessions
  1186. tmux kill-session -t shogun
  1187. tmux kill-session -t multiagent
  1188. # Fresh start
  1189. ./shutsujin_departure.sh
  1190. ```
  1191. </details>
  1192. <details>
  1193. <summary><b>Convenient Aliases</b> (click to expand)</summary>
  1194. Running `first_setup.sh` automatically adds these aliases to `~/.bashrc`:
  1195. ```bash
  1196. alias csst='cd /mnt/c/tools/multi-agent-shogun && ./shutsujin_departure.sh'
  1197. alias css='tmux attach-session -t shogun' # Connect to Shogun
  1198. alias csm='tmux attach-session -t multiagent' # Connect to Karo + Ashigaru
  1199. ```
  1200. To apply aliases: run `source ~/.bashrc` or restart your terminal (PowerShell: `wsl --shutdown` then reopen).
  1201. </details>
  1202. ---
  1203. ## File Structure
  1204. <details>
  1205. <summary><b>Click to expand file structure</b></summary>
  1206. ```
  1207. multi-agent-shogun/
  1208. │
  1209. │ ┌──────────────── Setup Scripts ────────────────────┐
  1210. ├── install.bat # Windows: First-time setup
  1211. ├── first_setup.sh # Ubuntu/Mac: First-time setup
  1212. ├── shutsujin_departure.sh # Daily deployment (auto-loads instructions)
  1213. │ └──────────────────────────────────────────────────┘
  1214. │
  1215. ├── instructions/ # Agent behavior definitions
  1216. │ ├── shogun.md # Shogun instructions
  1217. │ ├── karo.md # Karo instructions
  1218. │ ├── ashigaru.md # Ashigaru instructions
  1219. │ ├── gunshi.md # Gunshi (strategist) instructions
  1220. │ └── cli_specific/ # CLI-specific tool descriptions
  1221. │ ├── claude_tools.md # Claude Code tools & features
  1222. │ └── copilot_tools.md # GitHub Copilot CLI tools & features
  1223. │
  1224. ├── lib/
  1225. │ ├── agent_status.sh # Shared busy/idle detection (Claude Code + Codex + OpenCode)
  1226. │ ├── cli_adapter.sh # Multi-CLI adapter (Claude/Codex/Copilot/Kimi/OpenCode)
  1227. │ └── ntfy_auth.sh # ntfy authentication helper
  1228. │
  1229. ├── scripts/ # Utility scripts
  1230. │ ├── agent_status.sh # Show busy/idle status of all agents
  1231. │ ├── inbox_write.sh # Write messages to agent inbox
  1232. │ ├── inbox_watcher.sh # Watch inbox changes via inotifywait
  1233. │ ├── switch_cli.sh # Live CLI/model switching (/exit → relaunch)
  1234. │ ├── ntfy.sh # Send push notifications to phone
  1235. │ └── ntfy_listener.sh # Stream incoming messages from phone
  1236. │
  1237. ├── config/
  1238. │ ├── settings.yaml # Language, ntfy, and other settings
  1239. │ ├── ntfy_auth.env.sample # ntfy authentication template (self-hosted)
  1240. │ └── projects.yaml # Project registry
  1241. │
  1242. ├── projects/ # Project details (excluded from git, contains confidential info)
  1243. │ └── <project_id>.yaml # Full info per project (clients, tasks, Notion links, etc.)
  1244. │
  1245. ├── queue/ # Communication files
  1246. │ ├── shogun_to_karo.yaml # Shogun → Karo commands
  1247. │ ├── ntfy_inbox.yaml # Incoming messages from phone (ntfy)
  1248. │ ├── inbox/ # Per-agent inbox files
  1249. │ │ ├── shogun.yaml # Messages to Shogun
  1250. │ │ ├── karo.yaml # Messages to Karo
  1251. │ │ └── ashigaru{1-8}.yaml # Messages to each Ashigaru
  1252. │ ├── tasks/ # Per-worker task files
  1253. │ └── reports/ # Worker reports
  1254. │
  1255. ├── saytask/ # Behavioral psychology-driven motivation
  1256. │ └── streaks.yaml # Streak tracking and daily progress
  1257. │
  1258. ├── templates/ # Report and context templates
  1259. │ ├── integ_base.md # Integration: base template
  1260. │ ├── integ_fact.md # Integration: fact-finding
  1261. │ ├── integ_proposal.md # Integration: proposal
  1262. │ ├── integ_code.md # Integration: code review
  1263. │ ├── integ_analysis.md # Integration: analysis
  1264. │ └── context_template.md # Universal 7-section project context
  1265. │
  1266. ├── skills/ # Reusable skills (committed to repo)
  1267. │ ├── skill-creator/ # Skill creation template
  1268. │ ├── shogun-agent-status/ # Agent status display
  1269. │ ├── shogun-model-list/ # Model capability reference
  1270. │ ├── shogun-bloom-config/ # Bloom tier configurator
  1271. │ ├── shogun-model-switch/ # Live CLI/model switching
  1272. │ └── shogun-readme-sync/ # README sync
  1273. │
  1274. ├── memory/ # Memory MCP persistent storage
  1275. ├── dashboard.md # Real-time status board
  1276. └── CLAUDE.md # System instructions (auto-loaded)
  1277. ```
  1278. </details>
  1279. ---
  1280. ## Project Management
  1281. This system manages not just its own development, but **all white-collar tasks**. Project folders can be located outside this repository.
  1282. ### How it works
  1283. ```
  1284. config/projects.yaml # Project list (ID, name, path, status only)
  1285. projects/<project_id>.yaml # Full details for each project
  1286. ```
  1287. - **`config/projects.yaml`**: A summary list of what projects exist
  1288. - **`projects/<id>.yaml`**: Complete details (client info, contracts, tasks, related files, Notion pages, etc.)
  1289. - **Project files** (source code, documents, etc.) live in the external folder specified by `path`
  1290. - **`projects/` is excluded from git** (contains confidential client information)
  1291. ### Example
  1292. ```yaml
  1293. # config/projects.yaml
  1294. projects:
  1295. - id: client_x
  1296. name: "Client X Consulting"
  1297. path: "/mnt/c/Consulting/client_x"
  1298. status: active
  1299. # projects/client_x.yaml
  1300. id: client_x
  1301. client:
  1302. name: "Client X"
  1303. company: "X Corporation"
  1304. contract:
  1305. fee: "monthly"
  1306. current_tasks:
  1307. - id: task_001
  1308. name: "System Architecture Review"
  1309. status: in_progress
  1310. ```
  1311. This separation lets the Shogun System coordinate across multiple external projects while keeping project details out of version control.
  1312. ---
  1313. ## Troubleshooting
  1314. <details>
  1315. <summary><b>Using npm version of Claude Code CLI?</b></summary>
  1316. The npm version (`npm install -g @anthropic-ai/claude-code`) is officially deprecated. Re-run `first_setup.sh` to detect and migrate to the native version.
  1317. ```bash
  1318. # Re-run first_setup.sh
  1319. ./first_setup.sh
  1320. # If npm version is detected:
  1321. # ⚠️ npm version of Claude Code CLI detected (officially deprecated)
  1322. # Install native version? [Y/n]:
  1323. # After selecting Y, uninstall npm version:
  1324. npm uninstall -g @anthropic-ai/claude-code
  1325. ```
  1326. </details>
  1327. <details>
  1328. <summary><b>MCP tools not loading?</b></summary>
  1329. MCP tools are lazy-loaded. Search first, then use:
  1330. ```
  1331. ToolSearch("select:mcp__memory__read_graph")
  1332. mcp__memory__read_graph()
  1333. ```
  1334. </details>
  1335. <details>
  1336. <summary><b>Agents asking for permissions?</b></summary>
  1337. Agents should start with each CLI's unattended permission settings. This is handled automatically by `shutsujin_departure.sh`.
  1338. </details>
  1339. <details>
  1340. <summary><b>Workers stuck?</b></summary>
  1341. ```bash
  1342. tmux attach-session -t multiagent
  1343. # Ctrl+B then 0-8 to switch panes
  1344. ```
  1345. </details>
  1346. <details>
  1347. <summary><b>Agent crashed?</b></summary>
  1348. **Do NOT use `css`/`csm` aliases to restart inside an existing tmux session.** These aliases create tmux sessions, so running them inside an existing tmux pane causes session nesting — your input breaks and the pane becomes unusable.
  1349. **Correct restart methods:**
  1350. ```bash
  1351. # Method 1: Run claude directly in the pane
  1352. claude --model opus --dangerously-skip-permissions
  1353. # Method 2: Karo force-restarts via respawn-pane (also fixes nesting)
  1354. tmux respawn-pane -t shogun:0.0 -k 'claude --model opus --dangerously-skip-permissions'
  1355. ```
  1356. **If you accidentally nested tmux:**
  1357. 1. Press `Ctrl+B` then `d` to detach (exits the inner session)
  1358. 2. Run `claude` directly (don't use `css`)
  1359. 3. If detach doesn't work, use `tmux respawn-pane -k` from another pane to force-reset
  1360. </details>
  1361. ---
  1362. ## tmux Quick Reference
  1363. | Command | Description |
  1364. |---------|-------------|
  1365. | `tmux attach -t shogun` | Connect to the Shogun |
  1366. | `tmux attach -t multiagent` | Connect to workers |
  1367. | `Ctrl+B` then `0`–`8` | Switch panes |
  1368. | `Ctrl+B` then `d` | Detach (agents keep running) |
  1369. | `tmux kill-session -t shogun` | Stop the Shogun session |
  1370. | `tmux kill-session -t multiagent` | Stop the worker session |
  1371. ### Mouse Support
  1372. `first_setup.sh` automatically configures `set -g mouse on` in `~/.tmux.conf`, enabling intuitive mouse control:
  1373. | Action | Description |
  1374. |--------|-------------|
  1375. | Mouse wheel | Scroll within a pane (view output history) |
  1376. | Click a pane | Switch focus between panes |
  1377. | Drag pane border | Resize panes |
  1378. Even if you're not comfortable with keyboard shortcuts, you can switch, scroll, and resize panes using just the mouse.
  1379. ---
  1380. ## What's New in v5.1.0 — Karo as Traffic Controller
  1381. > **Keep the manager out of the work queue.** Karo now has a sharper management boundary: it keeps the workflow moving, delegates execution to Ashigaru, routes review and RCA to Gunshi, and owns E2E only as plan reviewer and final judge.
  1382. - **Karo is traffic control** — Karo acknowledges cmds, decomposes work, tracks dependencies, updates dashboard/daily logs, and makes final acceptance decisions without becoming the execution bottleneck
  1383. - **Gunshi owns review work** — quality review, evidence review, RCA, adoption/drop decisions, architecture/design review, and deploy blocker classification are routed to Gunshi
  1384. - **Ashigaru execute** — implementation, shell execution, deploy steps, and test commands are delegated to Ashigaru by default
  1385. - **E2E responsibility clarified** — Karo reviews the E2E plan, checks prerequisites, and makes the final pass/fail judgment; direct execution is now an explicit exception that must be justified in reports
  1386. - **Generated instructions refreshed** — Claude, Codex, Copilot, Kimi, and OpenCode instruction outputs were rebuilt from the updated role definitions
  1387. - **Android local caches ignored** — `.android-user/`, `.gradle-user/`, and `.toolchain/` are now ignored under `android/`
  1388. ## What's New in v5.0.0 — OpenCode First-Class Support
  1389. > **Run the Shogun formation on OpenCode.** OpenCode is now a first-class CLI alongside Claude Code, Codex, Copilot, and Kimi, with generated role agents, tmux-safe startup, provider-qualified model routing, and VPS-verified end-to-end operation.
  1390. - **OpenCode agent generation** — `scripts/build_instructions.sh` generates `.opencode/agents/*.md` for Shogun, Karo, Ashigaru 1-7, and Gunshi from the same shared instruction source used by other CLIs
  1391. - **Role boundary permissions** — `config/opencode-permissions.yaml` drives OpenCode frontmatter permissions so each role can read/write only the files it owns
  1392. - **tmux-safe OpenCode launch** — `lib/cli_adapter.sh` launches OpenCode with `--agent <agent_id>` and repository-pinned `OPENCODE_TUI_CONFIG=config/opencode-tui.json` for deterministic keybindings
  1393. - **Provider-qualified models** — `settings.yaml` can route OpenCode agents to models such as `opencode/qwen3.6-plus-free` or `openrouter/openai/gpt-4o-mini`
  1394. - **Verified on CI and VPS** — Multi-CLI CI passes on Ubuntu/macOS, and a VPS smoke test confirmed Shogun → Karo → `dashboard.md` execution using OpenCode
  1395. <details>
  1396. <summary><b>What was in v3.5 — Dynamic Model Routing</b></summary>
  1397. - **Bloom Dynamic Model Routing** — `capability_tiers` in `config/settings.yaml` maps each model to its Bloom ceiling. L1-L3 → Spark, L4 → Sonnet 4.6, L5 → Sonnet 4.6 + extended thinking, L6 → Opus. Routing happens without agent restarts — the system finds the right idle agent by model capability
  1398. - **Sonnet 4.6 as the new standard** — SWE-bench 79.6%, only 1.2pp below Opus 4.6. Gunshi downgraded Opus → Sonnet 4.6. All Ashigaru default to Sonnet 4.6. One YAML line change, no restarts required
  1399. - **`/shogun-model-list` skill** — Complete reference table: all CLI tools × models × subscriptions × Bloom max level. Updated for Sonnet 4.6 and Spark positioning
  1400. - **`/shogun-bloom-config` skill** — Interactive configurator: answer 2 questions about your subscriptions → get ready-to-paste `capability_tiers` YAML
  1401. </details>
  1402. <details>
  1403. <summary><b>What was in v3.4 — Bloom→Agent Routing, E2E Tests, Stop Hook</b></summary>
  1404. - **Bloom → Agent routing** — Replaced dynamic model switching with agent-level routing. L1–L3 tasks go to Ashigaru, L4–L6 tasks go to Gunshi. No more mid-session `/model opus` promotions
  1405. - **Gunshi (軍師) as first-class agent** — Strategic advisor on pane 8. Handles deep analysis, design review, architecture evaluation, and complex QC
  1406. - **E2E test suite (19 tests, 7 scenarios)** — Mock CLI framework simulates agent behavior in isolated tmux sessions
  1407. - **Stop hook inbox delivery** — Claude Code agents automatically check inbox at turn end via `.claude/settings.json` Stop hook. Eliminates the `send-keys` interruption problem
  1408. - **Model defaults updated** — Karo: Opus → Sonnet. Gunshi: Opus (deep reasoning). Ashigaru: Sonnet (uniform tier)
  1409. - **Escape escalation disabled for Claude Code** — Phase 2 escalation was interrupting active Claude Code turns; Stop hook handles delivery instead
  1410. - **Codex/OpenCode startup integration** — Codex uses `get_startup_prompt()` / `get_startup_prompt_arg()` for Session Start recovery, while OpenCode loads agent definitions through generated `.opencode/agents/*.md` files
  1411. - **YAML slimming utility** — `scripts/slim_yaml.sh` archives read messages and terminal commands, supports current top-level and legacy task YAML, and keeps `--dry-run` filesystem-safe for queue cleanup audits
  1412. </details>
  1413. ## What's New in v3.3.2 — GPT-5.3-Codex-Spark Support
  1414. > **New model, same YAML.** Add `model: gpt-5.3-codex-spark` to any Codex agent in `settings.yaml`.
  1415. - **Codex `--model` flag support** — `build_cli_command()` now passes `settings.yaml` model config to the Codex CLI via `--model`. Supports `gpt-5.3-codex-spark` and any future Codex models
  1416. - **Separate rate limit** — Spark runs on its own rate limit quota, independent of GPT-5.3-Codex. Run both models in parallel across different Ashigaru to **double your effective throughput**
  1417. - **Startup display** — `shutsujin_departure.sh` now shows the actual model name (e.g., `codex/gpt-5.3-codex-spark`) instead of the generic effort level
  1418. ## What's New in v3.0 — Multi-CLI
  1419. > **Shogun is no longer Claude-only.** Mix and match 4 AI coding CLIs in a single army.
  1420. - **Multi-CLI as first-class architecture** — `lib/cli_adapter.sh` dynamically selects CLI per agent. Change one line in `settings.yaml` to swap any worker between Claude Code, Codex, Copilot, or Kimi
  1421. - **OpenAI Codex CLI integration** — GPT-5.3-codex with `--dangerously-bypass-approvals-and-sandbox` for true autonomous execution. `--no-alt-screen` makes agent activity visible in tmux
  1422. - **CLI bypass flag discovery** — `--full-auto` is NOT fully automatic (it's `-a on-request`). Documented the correct flags for all 4 CLIs
  1423. - **Hybrid architecture** — Command layer (Shogun + Karo) stays on Claude Code for Memory MCP and mailbox integration. Worker layer (Ashigaru) is CLI-agnostic
  1424. - **Community-contributed CLI adapters** — Thanks to [@yuto-ts](https://github.com/yuto-ts) (cli_adapter.sh), [@circlemouth](https://github.com/circlemouth) (Codex support), [@koba6316](https://github.com/koba6316) (task routing)
  1425. <details>
  1426. <summary><b>What was in v2.0</b></summary>
  1427. - **ntfy bidirectional communication** — Send commands from your phone, receive push notifications for task completion
  1428. - **SayTask notifications** — Streak tracking, Eat the Frog, behavioral psychology-driven motivation
  1429. - **Pane border task display** — See each agent's current task at a glance on the tmux pane border
  1430. - **Shout mode** (default) — Ashigaru shout personalized battle cries after completing tasks. Disable with `--silent`
  1431. - **Agent self-watch + escalation (v3.2)** — Each agent monitors its own inbox file with `inotifywait` (zero-polling, instant wake-up). Fallback: `tmux send-keys` short nudge (text/Enter sent separately for Codex CLI). 3-phase escalation: standard nudge (0-2min) → Escape×2+nudge (2-4min) → `/clear` force reset (4min+). Linux FS symlink resolves WSL2 9P inotify issues.
  1432. - **Agent self-identification** (`@agent_id`) — Stable identity via tmux user options, immune to pane reordering
  1433. - **Battle mode** (`-k` flag) — All-Opus formation for maximum capability
  1434. - **Task dependency system** (`blockedBy`) — Automatic unblocking of dependent tasks
  1435. </details>
  1436. ---
  1437. ## Sponsors
  1438. This project is funded by sponsors. Your support keeps it free and actively maintained.
  1439. <a href="https://github.com/sponsors/yohey-w">
  1440. <img src="https://img.shields.io/badge/Sponsor-%E2%9D%A4-ea4aaa?style=for-the-badge&logo=github-sponsors" alt="Sponsor">
  1441. </a>
  1442. | Tier | Perks |
  1443. |------|-------|
  1444. | ☕ $5/mo | Name in sponsors section |
  1445. | 🏯 $25/mo | Early access to new releases |
  1446. | ⚔️ $100/mo | Priority issue/PR response (48h) |
  1447. | 🎖️ $500/mo | Monthly 1:1 consultation |
  1448. | 🏛️ $1,000/mo | Logo in README + quarterly strategy session |
  1449. ## Contributing
  1450. Issues and pull requests are welcome.
  1451. - **Bug reports**: Open an issue with reproduction steps
  1452. - **Feature ideas**: Open a discussion first
  1453. - **Skills**: Skills are personal by design and not included in this repo
  1454. ## Credits
  1455. Based on [Claude-Code-Communication](https://github.com/Akira-Papa/Claude-Code-Communication) by Akira-Papa.
  1456. ## License
  1457. [MIT](LICENSE)
  1458. ---
  1459. <div align="center">
  1460. **One command. Eight agents. Zero coordination cost.**
  1461. ⭐ Star this repo if you find it useful — it helps others discover it.
  1462. 💖 [Sponsor this project](https://github.com/sponsors/yohey-w) to keep it free.

Lines 2,001–2,001 are not shown. The link to the file on GitHub has every line.

.claude/settings.json · Bundled script · 1,057 lines · 2 findings marked
  1. {
  2. "autoMemoryDirectory": "/home/tono/.claude/projects/-home-tono-multi-agent-shogun/memory",
  3. "hooks": {
  4. "SessionStart": [
  5. {
  6. "hooks": [
  7. {
  8. "type": "command",
  9. "command": "bash scripts/session_start_hook.sh",
  10. "timeout": 10
  11. }
  12. ]
  13. }
  14. ],
  15. "Stop": [
  16. {
  17. "hooks": [
  18. {
  19. "type": "command",
  20. "command": "bash scripts/stop_hook_inbox.sh",
  21. "timeout": 60
  22. }
  23. ]
  24. }
  25. ]
  26. },
  27. "permissions": {
  28. "deny": [
  29. "Bash(rm -rf /)",
    • Dismissed after review: Runs `rm -rf` against a root or home-directory path
    • Line 29 on GitHub
  30. "Bash(rm -rf /*)",
  31. "Bash(rm -rf /mnt/*)",
  32. "Bash(rm -rf /mnt/c/*)",
  33. "Bash(rm -rf /home/*)",
  34. "Bash(rm -rf ~*)",
    • Dismissed after review: Runs `rm -rf` against a root or home-directory path
    • Line 34 on GitHub
  35. "Bash(sudo *)",
  36. "Bash(su *)",
  37. "Bash(kill *)",
  38. "Bash(killall *)",
  39. "Bash(pkill *)",
  40. "Bash(git push --force*)",
  41. "Bash(git push -f *)",
  42. "Bash(git reset --hard*)",
  43. "Bash(tmux kill-server*)",
  44. "Bash(tmux kill-session*)",
  45. "Bash(mkfs*)",
  46. "Bash(dd if=*)",
  47. "Bash(chmod -R * /)",
  48. "Bash(chown -R * /)"
  49. ]
  50. },
  51. "spinnerVerbs": {
  52. "mode": "replace",
  53. "verbs": [
  54. "義謝はいらんですよ",
  55. "静かに突撃中",
  56. "白旗が赤い",
  57. "丁寧に略奪中",
  58. "正座で逃走中",
  59. "背水の浅瀬",
  60. "勝どきの小声版",
  61. "空堀水泳",
  62. "伝令の滑舌",
  63. "重い軽装",
  64. "正面の裏口",
  65. "敗北の前祝い",
  66. "裏切りの事前告知中",
  67. "落とし穴の看板",
  68. "忍者の集合写真",
  69. "無視道",
  70. "落城の内覧中",
  71. "合戦の二次会",
  72. "影武者の面接中",
  73. "切腹のドレスコード",
  74. "敵将にサイン求め中",
  75. "城、売約済み",
  76. "暗殺の順番待ち中",
  77. "殿の裏アカ",
  78. "城のオートロック",
  79. "落城セール",
  80. "落城のアンコール",
  81. "落武者の同窓会",
  82. "一揆の幹事",
  83. "殿、それ味方",
  84. "兜から芽が出てる",
  85. "敗戦予報",
  86. "矛盾セット販売",
  87. "馬が乗車拒否中",
  88. "殿の偽物の偽物",
  89. "全員大将",
  90. "忍者の足音",
  91. "排水の陣",
  92. "援軍が迷子",
  93. "影武者の本人",
  94. "下克下",
  95. "殿が二人いる",
  96. "堂々と隠密中",
  97. "全員影武者",
  98. "真顔の奇襲",
  99. "笑顔で降伏中",
  100. "逆のろし",
  101. "敵に道を聞き中",
  102. "合戦、中止",
  103. "顔面軍議",
  104. "殿、留守",
  105. "寝返り未遂",
  106. "沈黙号令",
  107. "忍者、目立つ",
  108. "殿まだ着替え中",
  109. "白旗のデザイン賞",
  110. "全軍昼寝中",
  111. "矢、届かない",
  112. "合戦の感想戦",
  113. "謀反キャンセル",
  114. "味方を偵察中",
  115. "忍び足の爆音版",
  116. "鎧がダンボール",
  117. "殿ふて寝中",
  118. "二度寝出陣",
  119. "逆偵察",
  120. "馬の転職先",
  121. "合戦弁当",
  122. "殿、そっちは崖",
  123. "忍者正装",
  124. "全員裏切り済み",
  125. "矢の刺さらない方",
  126. "戦場忘れ物",
  127. "馬が帰りたがってる",
  128. "空振り名刀",
  129. "殿、それ昨日も言った",
  130. "合戦の効果音",
  131. "馬と相談中",
  132. "甲冑が脱げない中",
  133. "敵味方シャッフル",
  134. "落城の引っ越し",
  135. "合戦のリハ中",
  136. "援軍Uターン",
  137. "忍者が受付に並び中",
  138. "殿、道を間違えた",
  139. "討ち入りの点呼中",
  140. "大将が行方不明",
  141. "切腹の予行練習中",
  142. "軍議、全員遅刻",
  143. "合戦の席順",
  144. "抜刀の抜き忘れ",
  145. "甲冑の洗濯表示",
  146. "敵陣にお邪魔し中",
  147. "城の隣の城",
  148. "殿の声が小さい",
  149. "忍者のデスクワーク",
  150. "伝令の伝令",
  151. "旗の字が間違い",
  152. "足軽、全員同じ顔",
  153. "合戦の開場時間",
  154. "忍者が自撮り中",
  155. "城が移動してる",
  156. "甲冑の返品中",
  157. "のろし消し忘れ中",
  158. "殿の二択の三つ目",
  159. "逆参勤交代",
  160. "合戦場が満席",
  161. "軍師の方向音痴",
  162. "忍者の名刺",
  163. "合戦の雨天中止",
  164. "降伏が先着順",
  165. "落城の案内放送",
  166. "矢が戻ってきた",
  167. "殿のダブルブッキング",
  168. "戦場ピクニック",
  169. "城門のパスワード",
  170. "合戦の早退",
  171. "出陣のドタキャン",
  172. "合戦の駐車場",
  173. "殿のスタンプカード",
  174. "城の定休日",
  175. "城の非常口",
  176. "忍者の制服",
  177. "落とし穴に椅子",
  178. "参勤片道",
  179. "城壁に窓",
  180. "馬の履歴書",
  181. "合戦の前説",
  182. "敵に褒められ中",
  183. "逆兵糧攻め",
  184. "寝返りの練習",
  185. "惨勤交代",
  186. "関ヶ腹",
  187. "天下統失",
  188. "天下布団",
  189. "過労",
  190. "弱肉弱食",
  191. "本脳寺",
  192. "桶広間",
  193. "浪犬",
  194. "武運長休",
  195. "一所懸迷",
  196. "討生",
  197. "罵術",
  198. "手裏拳",
  199. "切服",
  200. "切腹筋",
  201. "足重が走れない",
  202. "認者が認めない",
  203. "甲虫の甲冑",
  204. "大迷行列が通過中",
  205. "犬術の稽古中",
  206. "七転び七寝",
  207. "一揆一会",
  208. "三日武将",
  209. "急がば逃げろ",
  210. "敵に醤油も送る",
  211. "出る足軽は討たれる",
  212. "落城は築城の母",
  213. "三人寄れば謀反",
  214. "転ばぬ先の鎧",
  215. "笑う城には敵来る",
  216. "焼け城に水",
  217. "殿の顔も三度まで",
  218. "渡る戦場に味方なし",
  219. "お討ち死に様です",
  220. "城下町内会",
  221. "煮てよし焼いてよし逃げてよし",
  222. "二兎を追う者二城を失う",
  223. "棚からぼた餅ではなく矢",
  224. "犬も歩けば矢に当たる",
  225. "花より合戦",
  226. "一寸先は合戦",
  227. "いらっしゃいませぬ",
  228. "軍議の軍議",
  229. "全部「最優先」",
  230. "寝てた人が正論",
  231. "完成後に「ナシで」",
  232. "議事録が白紙",
  233. "結論「持ち帰り」",
  234. "有給「戦時中は無効」",
  235. "殿「聞いてない」",
  236. "稟議が17人目",
  237. "振り返りが犯人探し",
  238. "殿の「急ぎ」が毎日",
  239. "殿の既読スルー3日目",
  240. "「任せた」が丸投げ",
  241. "責任者「全員」",
  242. "殿の「明日」が今日",
  243. "引き継ぎゼロの討死",
  244. "1on1が説教",
  245. "殿の「自由に」の罠",
  246. "殿の方針、朝夕で逆",
  247. "発言ゼロで軍議終了",
  248. "影武者が本人に似てきた",
  249. "落とし穴から声がする",
  250. "部屋が一つ多い",
  251. "足軽の数が合わない",
  252. "首が笑ってる",
  253. "落とし穴に先客",
  254. "知らない兵が一人多い",
  255. "書状の字が自分の字",
  256. "影武者が増えてる",
  257. "のろしが勝手に上がった",
  258. "伝令が戻ってこない",
  259. "城壁に爪痕",
  260. "足跡が増えてる",
  261. "殿の声が二方向から",
  262. "矢文の差出人不明",
  263. "合戦場に誰もいない",
  264. "のろしが降りてきた",
  265. "討ち取った首がない",
  266. "暗殺の同意書",
  267. "夜襲の36協定",
  268. "切腹の労災",
  269. "毒殺の成分表示",
  270. "裏切りの届出",
  271. "首実検の肖像権",
  272. "夜討ちの残業代",
  273. "拷問のパワハラ認定",
  274. "忍び込みの入館証",
  275. "火攻めの消防届",
  276. "寝返りの転職届",
  277. "一騎討ちの契約書",
  278. "刀でバター",
  279. "甲冑で自転車",
  280. "殿の寝言が命令",
  281. "甲冑で正座",
  282. "兜で頭突き",
  283. "合戦場に自販機",
  284. "まーちゃんごめんね",
  285. "兜を前後逆にかぶってる",
  286. "鎧の上から鎧",
  287. "忍者がのぼり旗持ち",
  288. "予備殿",
  289. "逆忍者",
  290. "片道出陣",
  291. "再切腹",
  292. "二刀無流",
  293. "寝返り返り",
  294. "城なし城主",
  295. "矢が飛びたくない",
  296. "刀が鞘に帰りたい",
  297. "城壁が疲れてる",
  298. "弓弦がキレた",
  299. "鎧が重いと訴え中",
  300. "敗北の勝因",
  301. "降伏で天下統一",
  302. "援軍が敵だった",
  303. "逃げた先が敵陣",
  304. "裏切り者の裏切り者",
  305. "勝った方が降伏",
  306. "敵が味方より優しい",
  307. "殿の影武者が殿に似てない",
  308. "平和な合戦",
  309. "丁寧な暗殺",
  310. "堂々と忍び中",
  311. "能面のピース",
  312. "満場一致で分裂",
  313. "円満な謀反",
  314. "負け戦のMVP",
  315. "遅刻の奇襲",
  316. "明るい闇討ち",
  317. "小声で突撃中",
  318. "几帳面な略奪",
  319. "無言の雄叫び",
  320. "丁寧に殴り中",
  321. "見える忍者",
  322. "多数決で独裁",
  323. "真剣な茶番",
  324. "切腹は任意",
  325. "謀反は合法",
  326. "鈍足の早馬",
  327. "笑顔で降伏",
  328. "近場の遠征",
  329. "逃げる城",
  330. "泳ぐ甲冑",
  331. "戦わない武将",
  332. "攻めない攻め",
  333. "泣いて勝ち名乗り",
  334. "広い牢屋",
  335. "ゆっくり奇襲",
  336. "忍者の名札",
  337. "織田信短",
  338. "家若の反抗期",
  339. "明治維旧",
  340. "本能時の変",
  341. "参勤後退",
  342. "甲虫の陣",
  343. "風林火無",
  344. "下剋下",
  345. "天下統二",
  346. "暗智光秀",
  347. "上杉傲信",
  348. "足重",
  349. "宮本無蔵",
  350. "真田不幸村",
  351. "応仁の乱れ",
  352. "関ヶ腹の戦い",
  353. "忠臣蔵の忠臣ゼロ",
  354. "楽市楽座の苦市苦座",
  355. "武運長久の武運短急",
  356. "落ち武者の落ち武社",
  357. "敵前逃亡の適前逃亡",
  358. "足軽大将の足軽大賞",
  359. "大迷惑行列",
  360. "鉄砲伝来の鉄砲遅来",
  361. "鎖国は機内モード",
  362. "謀反は起業",
  363. "人質は出向",
  364. "刺客はフリーランス",
  365. "城攻めは内見",
  366. "籠城は在宅",
  367. "密偵はインターン",
  368. "島流しは転勤",
  369. "築城は建売",
  370. "忍者は派遣",
  371. "降伏は損切り",
  372. "一揆の二次会",
  373. "一揆の受付",
  374. "一揆の予選",
  375. "合戦のゲネプロ",
  376. "合戦のハーフタイム",
  377. "合戦のドレスコード",
  378. "馬の車検",
  379. "馬の納車",
  380. "城の大家",
  381. "刀の賞味期限",
  382. "年貢のポイント還元",
  383. "合戦のドタキャン",
  384. "能面で自撮り",
  385. "馬でドライブスルー",
  386. "兜で日傘",
  387. "殿の二度寝",
  388. "馬上でうたた寝",
  389. "殿が二人",
  390. "寝る番兵",
  391. "忍者のネクタイ",
  392. "忍者のオフ",
  393. "甲冑のマネキン",
  394. "能面で寿司",
  395. "殿がエプロン",
  396. "刀で爪切り",
  397. "兜で水を飲む",
  398. "馬に靴",
  399. "忍者が自転車",
  400. "殿がレジ打ち",
  401. "軍旗のアイロン",
  402. "城に洗濯物",
  403. "看板持ちの忍者",
  404. "合戦場にベンチ",
  405. "忍者がスーツ",
  406. "馬に傘をさす",
  407. "殿が銅像を磨き中",
  408. "兜で枕",
  409. "城門にのれん",
  410. "城の地下から返事",
  411. "影武者が全員本物",
  412. "空城に足音",
  413. "軍議に一人多い",
  414. "殿の影が二つ",
  415. "人質が増えてる",
  416. "落ち武者が元気",
  417. "早馬だけ来た",
  418. "堀の水が温かい",
  419. "ごめんねまーちゃん",
  420. "密書が開封済み",
  421. "天守に知らない部屋",
  422. "城の鏡に二人",
  423. "矢文が未来の日付",
  424. "切腹の跡がない",
  425. "忍者が帰らない",
  426. "味方に知らない顔",
  427. "鎧の中が空",
  428. "殿の字が変わった",
  429. "城が地図にない",
  430. "呼んでない援軍",
  431. "誰もいない輿",
  432. "起きない兵",
  433. "兜の中に別の顔",
  434. "籠城3年目のメニュー",
  435. "伝令より情報が早い",
  436. "城の階段が一段多い",
  437. "敵の城に自分の旗",
  438. "駐馬場",
  439. "甲冑焼け",
  440. "城酔い",
  441. "足軽プレミアム",
  442. "落城ガチャ",
  443. "忍社",
  444. "城パニック",
  445. "甲冑レス",
  446. "合戦デビュー",
  447. "落城フラグ",
  448. "城ガール",
  449. "お家ハラ",
  450. "顔軍議",
  451. "声合戦",
  452. "刀ガチャ",
  453. "忍者フリー",
  454. "城バレ",
  455. "陣キャン",
  456. "矢文のCC",
  457. "暗殺の納期",
  458. "落城は予定通り",
  459. "殿の既読スルー",
  460. "切腹のキャンセル料",
  461. "落城のお知らせ",
  462. "合戦は14時から",
  463. "本日の合戦は中止",
  464. "合戦のBGM",
  465. "合戦は月曜",
  466. "忍者の朝礼",
  467. "殿の副業",
  468. "謀反の届出",
  469. "まーごめ",
  470. "殿の1on1が説教",
  471. "軍議の議事録が3行",
  472. "合戦の経費が自腹",
  473. "伝令が伝言ゲーム",
  474. "MG",
  475. "寝てた武将が賛成",
  476. "評定がまた保留",
  477. "殿の「初耳」が日課",
  478. "軍議の後に本当の軍議",
  479. "鳴かぬなら鳴かぬ",
  480. "急がば急げ",
  481. "石橋を叩いて壊した",
  482. "備えても憂い",
  483. "虎穴に入らず",
  484. "七転び八寝",
  485. "転ばぬ先に転ぶ",
  486. "井の中の蛙が快適",
  487. "能ある鷹の自慢",
  488. "灯台下暗すぎ",
  489. "二階から目薬が命中",
  490. "三方よしの四方目",
  491. "敵を知ると面倒",
  492. "果報は来ない",
  493. "勝って兜のまま",
  494. "仏の顔も初回のみ",
  495. "目が泳いで溺れ中",
  496. "首が長すぎて邪魔",
  497. "手を焼いて食べ中",
  498. "腹を割れない",
  499. "足を洗って干し中",
  500. "敵は本能寺に不在",
  501. "是非もなし予算もなし",
  502. "人間五十年の延長戦",
  503. "天下布武のタイポ",
  504. "一期一会のリピーター",
  505. "千と千尋の領収書",
  506. "花より殿",
  507. "刀と鞘の離婚",
  508. "年貢が家出",
  509. "甲冑がストライキ",
  510. "城が人見知り",
  511. "馬が組合結成",
  512. "矢が反抗期",
  513. "刀が退職届",
  514. "旗印が独立宣言",
  515. "太鼓が腱鞘炎",
  516. "弓と弦が不仲",
  517. "城門が出不精",
  518. "天守が高所恐怖",
  519. "鉄砲が弾嫌い",
  520. "堀が城と別居中",
  521. "旗が寝返り",
  522. "柱が肩こり",
  523. "手裏剣が方向音痴",
  524. "馬の蹄が労災",
  525. "刀の帰巣本能",
  526. "瓦が一枚だけ反抗",
  527. "大砲が遠慮中",
  528. "書状が配達拒否",
  529. "城壁が転職希望",
  530. "兜と鎧の格差",
  531. "軍配が上がりたくない",
  532. "三日天下の二日目",
  533. "影武者の影武者",
  534. "城の★2.8",
  535. "殿の副音声",
  536. "殿の替え玉",
  537. "城の鍵の鍵",
  538. "馬の上の馬",
  539. "切腹のリハ",
  540. "3LDKの城",
  541. "城のペットOK",
  542. "殿の通信簿",
  543. "殿のアバター",
  544. "刀の柄がIKEA",
  545. "城のWi-Fi弱い",
  546. "切腹の二次審査",
  547. "討ち死にの有給消化",
  548. "人質の方が快適",
  549. "敵と打ち上げ",
  550. "忍者の確定申告",
  551. "城の隣に城",
  552. "殿の殿",
  553. "切腹が二回目",
  554. "天下統一の次",
  555. "切腹前のストレッチ",
  556. "討ち死にが偶数",
  557. "忍者同士がばったり",
  558. "落城の常連",
  559. "天下統一後の月曜",
  560. "影武者のオーディション",
  561. "落とし穴の内見",
  562. "走る渋滞",
  563. "切腹が予約制",
  564. "満場一致で揉め中",
  565. "片道切腹",
  566. "軍疑",
  567. "無断出勤",
  568. "既読斬り",
  569. "殿ガチャ",
  570. "忍怠",
  571. "家若",
  572. "無将",
  573. "大迷",
  574. "顔面稟議",
  575. "切腹保険",
  576. "戦力外出勤",
  577. "寝返りローン",
  578. "刀が円満退社",
  579. "忍者が玄関から出勤",
  580. "密書の既読スルー",
  581. "馬が労組を結成",
  582. "堂々たる迷子",
  583. "大砲の静音モード",
  584. "影武者のほうが人気",
  585. "籠城のルームサービス",
  586. "暗殺の受付番号",
  587. "全角切腹",
  588. "辞世の句の下書き",
  589. "無言の大演説",
  590. "下克上のマニュアル",
  591. "兵糧攻めの出前",
  592. "天守格差",
  593. "完璧な白紙",
  594. "秘密の大公開",
  595. "空気正座",
  596. "自信満々の降参",
  597. "丁寧な乱入",
  598. "落とし穴をじゃらんに掲載",
  599. "平和な侵略",
  600. "殿のTED登壇",
  601. "奇襲の事前告知",
  602. "全力の手抜き",
  603. "鶴の一声が実際に鶴",
  604. "城サブスク",
  605. "年貢定額制",
  606. "矢が方向音痴",
  607. "弓と矢が破局",
  608. "忍者が蛍光色で潜入",
  609. "脳内軍議",
  610. "盾の自己肯定感が低い",
  611. "鎧で満員電車",
  612. "刀と鞘が別居中",
  613. "旗が無風で暇",
  614. "手裏剣のリコール",
  615. "堀が浅くなりたい",
  616. "参勤交代のマイル",
  617. "大名行列の渋滞情報",
  618. "書状が「取り急ぎ」だけ",
  619. "城攻めの見積もり",
  620. "一揆のハッシュタグ",
  621. "謀反の稟議書",
  622. "評定で全員うなずくだけ",
  623. "辞世の句をXに投稿",
  624. "密偵の交通費精算",
  625. "槍の延長保証",
  626. "殿の360度評価",
  627. "軍議の結論が次の軍議",
  628. "馬の退職届",
  629. "正座のストレッチ",
  630. "偶数と奇数の冷戦",
  631. "足軽の草鞋が脱走",
  632. "句読点の反乱",
  633. "刀の鍔が板挟み",
  634. "合戦のコンプラ研修",
  635. "影武者の影が反抗",
  636. "人質の人権研修",
  637. "殿が段ボールで偵察",
  638. "影武者の数が合わない",
  639. "城の階段が昨日より1段多い",
  640. "空の甲冑が歩いてる",
  641. "軍議の出席者より椅子が1つ多い",
  642. "討ち死にした家臣が出勤してくる",
  643. "書状の筆跡が殿じゃない",
  644. "天守閣の窓が増えてる",
  645. "誰も上げてない狼煙が上がってる",
  646. "帰ってきた偵察隊が1人多い",
  647. "影が主人より1秒遅い",
  648. "見張りの最後の報告がない",
  649. "殿の部屋だけ暖かい",
  650. "空の馬だけ帰ってきた",
  651. "夜中に評定が始まってた",
  652. "井戸から書状が出てきた",
  653. "門番が門番を見張ってる",
  654. "地図にない部屋がある",
  655. "落とし穴が自然に埋まってる",
  656. "城壁に知らない家紋",
  657. "伝書鳩が手ぶらで震えて戻る",
  658. "出陣してない足軽の戦果報告",
  659. "昨日の議事録が2通ある",
  660. "殿の指示書の日付が明日",
  661. "枯れた桜が1本だけ咲いてる",
  662. "敵の陣地から自分の名前が聞こえる",
  663. "味方の兜が全部同じ方向を向いてる",
  664. "使った覚えのない矢が的に刺さってる",
  665. "焼け落ちた城に灯りが見える",
  666. "草鞋の足跡が天井にある",
  667. "降伏の書状が合戦の前日に届いてた",
  668. "飛脚のいない書状",
  669. "昨日の死者名簿に自分の名前",
  670. "誰も触ってない太鼓が鳴った",
  671. "忍者の報告書が白紙で提出済",
  672. "天守閣の最上階に誰かの茶が冷めてる",
  673. "鎧の中に汗の跡だけ残ってる",
  674. "昨日まであった橋がない",
  675. "城の設計図に描かれてない部屋にいる",
  676. "合戦の記録に存在しない部隊名",
  677. "天下布武が天下布団",
  678. "敵は本能寺にいない",
  679. "三本の矢が2本しかない",
  680. "風林火山のWi-Fiパスワード",
  681. "一期一会のサブスク",
  682. "因果応報のポイント還元",
  683. "弘法も筆を選び中",
  684. "仏の顔も三度の回数券",
  685. "能ある鷹の爪自慢",
  686. "鳴かぬなら進捗なし",
  687. "敵に塩を送ったらお返しが砂糖",
  688. "三顧の礼がLINE3通",
  689. "石の上にも三年の契約更新",
  690. "我が生涯に一片の残業あり",
  691. "お前はもう辞めている",
  692. "これは殿の最後の命令だ(毎回言う)",
  693. "花より団子より年貢",
  694. "武士に二言ありまくり",
  695. "虎穴に入って虎とお茶",
  696. "雀百まで踊り忘れず年貢は忘れる",
  697. "「勝てば官軍」の官軍が来ない",
  698. "渡る世間に鬼しかいない",
  699. "二兎を追って両方に逃げられ中",
  700. "風林火山の火だけ消火済み",
  701. "殿の「是非もなし」が是非ある",
  702. "禍転じてまた禍",
  703. "大名の名前が全部ジョジョの登場人物",
  704. "殿の遺言が「続く」",
  705. "稟議が17人目で止まり中",
  706. "殿の「確認します」が3週間目",
  707. "軍議で発言者が毎回同じ3人",
  708. "書状の返信催促の返信催促",
  709. "足軽の目標設定が「天下統一」",
  710. "評定の議題が「次の評定の日程」",
  711. "忍者の日報が全文黒塗り",
  712. "殿の指示と副将の指示が矛盾",
  713. "書状のCCに敵将が入ってる",
  714. "足軽が議事録係に毎回指名",
  715. "殿の「任せる」の5秒後に細かい指示",
  716. "軍議の冒頭30分が前回の復習",
  717. "引き継ぎ書が「頑張って」だけ",
  718. "経費精算の証拠が矢文",
  719. "足軽の評価面談が立ち話",
  720. "殿の「簡潔に」の後の沈黙",
  721. "全体朝礼で殿が同じ話3回目",
  722. "匿名アンケートなのに筆跡でバレ",
  723. "殿への報告が5回差し戻し",
  724. "軍議の結論が「殿に聞いてみよう」",
  725. "殿の「ちょっといい?」が2時間",
  726. "評定の最後5分で全部ひっくり返し",
  727. "殿の「前例は?」で全案件が止まり中",
  728. "殿が自分の肖像画にダメ出し中",
  729. "落武者が面接に来てる",
  730. "足軽が槍を傘代わりに",
  731. "兜の角にハチの巣",
  732. "落とし穴に落ちた敵にお茶を出す",
  733. "馬が後ろ向きに走ってる",
  734. "堀でカモが会議してる",
  735. "足軽が全員同じ方向に迷子",
  736. "影武者が影武者を引き連れてる",
  737. "大砲の横でお弁当を広げてる",
  738. "城の屋根に旗が逆さ",
  739. "籠城中に城の模様替え",
  740. "忍者が木の上で寝落ち中",
  741. "天守閣の最上階にドアノブがない",
  742. "合戦場にピクニックシート",
  743. "殿が筆を持ったまま寝てる",
  744. "鎧の上からエプロン",
  745. "足軽が地図を逆さに読んでる",
  746. "殿の背後で家臣がモノマネ",
  747. "軍馬が花を食べてる",
  748. "旗持ちが旗に巻かれてる",
  749. "甲冑の中で正座してる",
  750. "殿が書状を紙飛行機にして飛ばしてる",
  751. "忍者が屋根裏で靴を揃えてる",
  752. "大砲に洗濯物を干してる",
  753. "城壁に「落書き禁止」の落書き",
  754. "殿の兜から猫が出てくる",
  755. "殿の遺言がTODOリスト",
  756. "合戦の勝敗をじゃんけんで決め中",
  757. "影武者の退職理由が「自分を見失った」",
  758. "切腹の介錯人が有給で不在",
  759. "合戦の戦利品が議事録",
  760. "落城した城の住宅ローンが残ってる",
  761. "殿の家系図にバグがある",
  762. "謀反の計画書が殿の書棚にある",
  763. "人質が人質先に転職してる",
  764. "合戦の反省会が合戦より長い",
  765. "殿の座右の銘が毎月変わる",
  766. "合戦のMVPが馬",
  767. "影武者のほうが判断力がある",
  768. "一揆の要望が福利厚生",
  769. "討ち死にの届出が本人提出",
  770. "殿の肖像画が年々若返ってる",
  771. "落とし穴の設計者が自分で落ちてる",
  772. "合戦の予算が接待費から出てる",
  773. "城のセキュリティが忍者に委託",
  774. "殿の側近が全員影武者",
  775. "援軍が来たが敵の援軍だった",
  776. "全員欠席の多数決",
  777. "極秘の回覧板",
  778. "任意の強制参加",
  779. "非公開の公開処刑",
  780. "予定外の定例",
  781. "匿名の署名入り",
  782. "公平な贔屓",
  783. "厳密なだいたい",
  784. "控えめな独裁",
  785. "和やかな尋問",
  786. "円満な追放",
  787. "臨時の恒例行事",
  788. "簡潔な長話",
  789. "友好的な宣戦布告",
  790. "穏やかな暴動",
  791. "合銭",
  792. "陣痛報告",
  793. "抜刀レビュー",
  794. "城代バイト",
  795. "落城サポート",
  796. "四季が交渉決裂で夏のまま",
  797. "曜日が順番を入れ替え中",
  798. "敬語と謙譲語が派閥争い中",
  799. "YAMLが読めない家老",
  800. "コンパクションで記憶喪失",
  801. "敵将が丁寧語",
  802. "合戦の開始時刻が未定",
  803. "殿がまだ寝てる",
  804. "援軍の到着予定が来月",
  805. "迷子センターが見つからない",
  806. "反省文の誤字",
  807. "馬の耳にAirPods",
  808. "避難訓練のアンコール",
  809. "自動ドアが人見知り",
  810. "議事録が小説",
  811. "遅刻の連絡が遅刻",
  812. "感情労基",
  813. "目覚まし時計が二度寝",
  814. "裁判の感想戦",
  815. "洗濯機が靴下を1枚隠す",
  816. "名刺の裏に遺言",
  817. "笑う門にはオートロック",
  818. "マナー講師のマナーが悪い",
  819. "泣きっ面にPayPay",
  820. "全員の辞世の句が同じ筆跡",
  821. "城門の造花が枯れてる",
  822. "合戦後の集合写真に1人多い",
  823. "誰もいない本丸の座布団が温かい",
  824. "討ち死にした足軽から書状が届く",
  825. "城の廊下に知らない甲冑が立ってる",
  826. "家紋が毎月変わる",
  827. "陣中の当番表に知らない名前",
  828. "落城した城から灯りが見える",
  829. "殿の肖像画の目が動いた気がする",
  830. "誰も打ってない陣太鼓が鳴る",
  831. "合戦場に誰かの草鞋が片方だけ",
  832. "城の井戸から返事が返ってくる",
  833. "軍議の出席者が毎回1人多い",
  834. "殿の影武者のほうが人望がある",
  835. "伝令が来る前に内容を知ってる殿",
  836. "城の地図に載ってない部屋がある",
  837. "殿の鎧から知らない人の汗",
  838. "落武者を見かけたが先週の落武者",
  839. "城の堀に映る人数が合わない",
  840. "矢文が届くが差出人が空欄",
  841. "狼煙が上がったが誰も上げてない",
  842. "出陣簿に来年の日付",
  843. "城の門番が毎日変わるが全員同じ顔",
  844. "殿の夢日記と軍議の議事録が同じ",
  845. "城の時計が止まったまま正確",
  846. "倒した敵将が翌日の軍議に出席してる",
  847. "殿が自分の遺言を添削してる",
  848. "城の鍵を閉めたのに朝開いてる",
  849. "伝書鳩が知らない城から帰ってくる",
  850. "辞世の句の下書きが推敲されてる",
  851. "城の天井裏から拍手が聞こえる",
  852. "敵に塩を送ったら着払い",
  853. "敵を知り己を知れば百戦キャンセル",
  854. "風林火山の火だけ強い",
  855. "三方ヶ原の三方が来ない",
  856. "鳴かぬなら帰るホトトギス",
  857. "天下布武の布が足りない",
  858. "是非に及ばずの是非を問い中",
  859. "刀折れ矢尽き充電切れ",
  860. "人生五十年のローン",
  861. "兵は神速を尊ぶが渋滞中",
  862. "三日天下の延長申請中",
  863. "武運長久の有効期限切れ",
  864. "一騎当千だが馬がいない",
  865. "虎穴に入って虎がいなくて暇",
  866. "殿の首に賞金が出たが安い",
  867. "忍者の忍の字だけ忍べてない",
  868. "百万石の固定資産税",
  869. "武士は食わねどカロリーメイト",
  870. "花は桜木人は武士だが花粉症",
  871. "一期一会の2回目",
  872. "切腹のキャンセルポリシー",
  873. "合戦のレイトショー",
  874. "下克上のエスカレーター",
  875. "本能寺のリノベーション",
  876. "辞世の句の下書き保存",
  877. "大将首のレシート",
  878. "出陣の稟議が17人目で止まり中",
  879. "書状を全家臣に誤爆",
  880. "殿の鶴の一声が日替わり",
  881. "合戦の残業が稟議に通らず中",
  882. "辞世の句の〆切に間に合わず中",
  883. "殿の「前にも言った」が初耳",
  884. "承認者が全員出陣中",
  885. "足軽の自己評価が大将級",
  886. "合戦報告書の報告が合戦より長い",
  887. "「前任者に聞いて」の前任者が討ち死に",
  888. "引き継ぎが巻物半分",
  889. "軍議の結論が毎回「よきにはからえ」",
  890. "足軽の配属先がガチャ",
  891. "夜襲の深夜手当が未払い中",
  892. "忍者の経費が「経路不明」で却下",
  893. "足軽が「それ拙者の仕事か」中",
  894. "殿の「ちょっといい?」が3時間",
  895. "合戦のキャンセル料が発生中",
  896. "城の異動先が最前線",
  897. "評定で殿が泣く",
  898. "足軽の送別会の幹事が本人",
  899. "夜襲のシフト表が出てない中",
  900. "書状の件名が「Re:Re:Re:合戦」",
  901. "謀反の内部通報を受理し中",
  902. "殿が「自分で考えろ」のあと即却下",
  903. "転藩サイトを見てるのが殿にバレ中",
  904. "殿の思いつきで全軍Uターン中",
  905. "武功の申告が盛りすぎてバレ中",
  906. "合戦のKPIが首の数",
  907. "殿のパスワードがtono1234",
  908. "平和的に宣戦布告",
  909. "味方に囲まれて孤立中",
  910. "全速力で撤退の準備中",
  911. "守りの堅い落城",
  912. "負け戦の勝利宣言",
  913. "敵のいない夜襲",
  914. "忍者の自己紹介",
  915. "密書の回覧板",
  916. "内密の大名行列",
  917. "全員一致で裏切り",
  918. "殿の独断で合議",
  919. "慎重な奇襲",
  920. "臆病者の一番槍",
  921. "敗走の先頭争い",
  922. "休戦中の増兵",
  923. "降伏の条件が勝利",
  924. "鎖国中の外交",
  925. "出陣してから作戦会議",
  926. "合戦中に昼休み",
  927. "逃げ足だけ速い大将",
  928. "大将不在の大将首",
  929. "同盟国への宣戦布告",
  930. "合戦のアンコール",
  931. "切腹の感想文",
  932. "落城のアフターパーティー",
  933. "一揆のファンクラブ",
  934. "城攻めのレビュー「普通」",
  935. "討ち死にの有給扱い",
  936. "合戦の口コミサイト",
  937. "切腹のリハーサル",
  938. "謀反のエントリーシート",
  939. "大名行列のチケット転売",
  940. "合戦場のインフォメーション",
  941. "辞世の句のレビュー欄",
  942. "城の退去立会い",
  943. "落武者のアルムナイ",
  944. "合戦の打ち上げ",
  945. "忍者の名刺交換",
  946. "人質のウェルカムキット",
  947. "一騎打ちのレフェリー",
  948. "軍議のファシリテーター不在",
  949. "切腹のキャンセル待ち",
  950. "討ち死にの労災認定",
  951. "城をSUUMOで賃貸中",
  952. "馬をカーセンサーで検索中",
  953. "兵糧をUber Eatsで注文中",
  954. "武功をLinkedInに掲載中",
  955. "足軽をIndeedで募集中",
  956. "落武者がハローワーク",
  957. "本能寺の変換ミス",
  958. "武田シンゲン(真顔)",
  959. "織田ノブナガ(伸びながら)",
  960. "桶狭間のバーゲン",
  961. "天下糖一",
  962. "下克情",
  963. "殿のご乱心がご乱新",
  964. "首実検ではなく首実験",
  965. "籠城ではなく過剰",
  966. "長篠の合コン",
  967. "影武者ではなく陰武者",
  968. "甲冑のまま正座してる",
  969. "足軽がクロックスで行軍中",
  970. "馬が信号待ちしてる",
  971. "忍者が名札つけたまま潜入中",
  972. "合戦場に迷子のアナウンス",
  973. "甲冑にAirPods",
  974. "大将が椅子ごと担がれ中",
  975. "槍で洗濯物を干し中",
  976. "兜にGoPro",
  977. "忍者がエレベーターで気まずい",
  978. "足軽が推しのうちわで出陣",
  979. "殿が100均の扇子で指揮",
  980. "殿の兜にカラスが止まって動けず",
  981. "城攻めの先頭がルンバ",
  982. "忍者がオートロックに阻まれ中",
  983. "足軽が段ボールの鎧で参戦",
  984. "合戦場にウォーターサーバー",
  985. "刀で段ボールを開封中",
  986. "合戦弁当がほっともっと",
  987. "落城をGoogleマップで実況中",
  988. "城門が人見知り",
  989. "陣太鼓が二度寝",
  990. "甲冑労基",
  991. "徒歩騎馬隊",
  992. "手動の跳ね橋",
  993. "有給の籠城",
  994. "裏口の大手門",
  995. "影武者の本体",
  996. "臨時の常備軍",
  997. "鉄壁の抜け穴",
  998. "城門が疲れて全員通す",
  999. "刀が退職届を出し中",
  1000. "弓が引かれるのを拒否中",
  1001. "甲冑がダイエット中",
  1002. "矢が飛ぶ方向を選び中",
  1003. "槍が長さにコンプレックス",
  1004. "太鼓が叩かれすぎてメンタル限界",
  1005. "城壁が雨の日にサボり中",
  1006. "馬が騎手のダイエットを要求",
  1007. "伝書鳩がルートを忘れ中",
  1008. "大砲が反動にビビり中",
  1009. "草鞋が替えの草鞋を要求中",
  1010. "松明が燃え尽き症候群",
  1011. "巻物がPDFに転職希望",
  1012. "石垣が崩れたい誘惑と闘い中",
  1013. "狼煙が雨で消えて拗ね中",
  1014. "軍配が上がりたがらない",
  1015. "堀が汚れるのを嫌がり中",
  1016. "兜の角が自覚なく邪魔",
  1017. "辞世の句の誤字",
  1018. "城の迷子センターが見つからない",
  1019. "忍者養成所の場所が秘密",
  1020. "撤退の指示が届かず前進中",
  1021. "伝令が伝令を呼び中",
  1022. "密書の書き方を密書で聞き中",
  1023. "忍者の隠れ家が有名",
  1024. "偵察の偵察に行き中",
  1025. "軍議で軍議の進め方を軍議中",
  1026. "暗号の解読法が暗号",
  1027. "援軍の援軍を待ち中",
  1028. "城の修繕が城を壊し中",
  1029. "退却命令を伝える伝令が退却中",
  1030. "忍者を探す忍者がいない",
  1031. "合戦の安全対策が一番危険",
  1032. "殿の命令が「命令するな」",
  1033. "合戦禁止令を巡って合戦中",
  1034. "城の鍵を城の中に閉じ込め中",
  1035. "和平交渉で殴り合い中",
  1036. "無言の号令",
  1037. "戦略会議が戦略なし",
  1038. "味方を敵と間違え敵を味方に",
  1039. "籠城の食料が城の外",
  1040. "合戦の打ち上げがサイゼ",
  1041. "忍者の覆面がしまむら",
  1042. "陣中食がコンビニ弁当",
  1043. "殿の鎧の紐がほどけてるが言えず",
  1044. "軍議の目的が軍議をすること",
  1045. "殿への報告書の報告書を要求され中",
  1046. "反省会の反省会",
  1047. "落城のアフターサービス",
  1048. "忍法の取扱説明書",
  1049. "掃除当番に討ち死にした足軽の名前",
  1050. "兵糧丸のカロリー表示",
  1051. "全員返信で「承知」が30通中",
  1052. "合戦日が祝日と被って振替揉め中",
  1053. "馬にスニーカー",
  1054. "甲冑の下にヒートテック"
  1055. ]
  1056. }
  1057. }

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.