### Setup Completion Report

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-d-stackchan.md

A confirmation message detailing the Stack-chan integration setup, including connection details, hardware, TTS, UI link, and log file location. Provides instructions to start and stop the service.

```bash
xangi-stackchan integration complete (Mode D)

Connected xangi: http://127.0.0.1:18888
Hardware: <K151-R / CoreS3 standalone / AtomS3R + Voice Base>
TTS: <piper / voicevox / none>
Settings UI: http://127.0.0.1:7897/
Log: tail -f /tmp/xangi-stackchan.log

First test: Talk to xangi -> Stack-chan speaks the response
To stop: pkill -f xangi-stackchan
```

--------------------------------

### Install gogcli by downloading binary

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Manually install gogcli by downloading the binary from the releases page and extracting it to a directory in your PATH.

```bash
# バイナリ直接ダウンロード（リリースページ: https://github.com/steipete/gogcli/releases）
mkdir -p ~/bin
curl -L https://github.com/steipete/gogcli/releases/latest/download/gogcli_<version>_<os>_<arch>.tar.gz | tar xz -C ~/bin gog
```

--------------------------------

### Greeting Message

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/BOOTSTRAP.md

Display a friendly greeting to the user to start the setup process.

```text
はじめまして！あなた専用のAIアシスタントを作りましょう。
いくつか質問させてください。3分くらいで終わります。
```

--------------------------------

### Setup Completion Message

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/BOOTSTRAP.md

Display a confirmation message upon successful setup, including the AI's name and suggested next actions.

```text
セットアップ完了！

[AIの名前]として、これからよろしくお願いします。
次回からは自動的にあなたの設定を読み込みます。

まずは何か試してみませんか？
- 「今日の日記を書いて」
- 「最新のテックニュースを教えて」
- 「スライドを作って」
```

--------------------------------

### Enable and Start Xangi Systemd Service

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

Enables the Xangi systemd service to start on boot and starts it immediately.

```bash
sudo systemctl enable --now xangi.service
```

--------------------------------

### Check Prerequisites

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

Verify that Docker, Git, and sufficient disk space are available before proceeding with the setup.

```bash
docker --version && docker compose version
git --version
df -h ~ | tail -1   # 空きが 10GB 以上
```

--------------------------------

### Install gogcli using Homebrew

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Install the gogcli tool on macOS or Linux using the Homebrew package manager.

```bash
# macOS / Linux (Homebrew)
brew install gogcli
```

--------------------------------

### Install gogcli on Arch Linux

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Install the gogcli tool on Arch Linux using the yay AUR helper.

```bash
# Arch Linux
yay -S gogcli
```

--------------------------------

### Setup xangi-stackchan Bridge and TTS

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-d-stackchan.md

Set up the xangi-stackchan bridge on your PC using uv and install the piper-plus TTS engine with the Tsukuyomi-chan model. This enables text-to-speech capabilities for Stack-chan.

```bash
cd xangi-stackchan
uv sync
./scripts/setup_piper.sh
```

--------------------------------

### Caddyfile Example for Reverse Proxy with Basic Auth

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

An example Caddyfile configuration to set up a reverse proxy for Xangi, including SSL termination and basic authentication for access control.

```caddy
# Caddyfile 例
xangi.example.com {
  basicauth {
    user $2a$14$...
  }
  reverse_proxy 127.0.0.1:18888
}
```

--------------------------------

### Example Meal and Exercise Inputs

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/health-advisor/SKILL.md

Examples of user inputs that trigger meal and exercise logging, including text descriptions and commands.

```text
昼はカレーうどん食べた
散歩してきた、5000歩
サウナ行ってきた、2セット
週次ヘルスレポート
健康チェック
```

--------------------------------

### Set up Notion API Key

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/notion-manager/SKILL.md

Configure the Notion API key by creating a directory and echoing the API key into a file. This is a one-time setup step.

```bash
mkdir -p ~/.config/notion
echo "ntn_xxxxx" > ~/.config/notion/api_key
```

--------------------------------

### Setup Completion Report

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

A confirmation message indicating that Xangi has been successfully started in Discord mode, including connection details and next steps.

```text
xangi 起動完了（Discord モード）

接続先: Discord サーバー <名前>
Bot: <Bot 名>
Web UI (任意): http://localhost:18888
ワークスペース: ~/xangi/xangi/ai-assistant-workspace

最初に試して: Bot に DM またはメンションで「こんにちは」
停止: docker compose down
```

--------------------------------

### Add Custom RSS Feed Example

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/tech-news-curation/SKILL.md

Example command to fetch content from a custom RSS feed. Add your preferred feeds here for customized news curation.

```bash
# 追加例
curl -s "https://example.com/feed.xml" | head -200
```

--------------------------------

### Clone Xangi and Copy Environment File

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

Clone the Xangi repository and copy the example environment file to create a new configuration file.

```bash
mkdir -p ~/xangi && cd ~/xangi
git clone https://github.com/karaage0703/xangi.git
cd xangi
cp .env.example .env
```

--------------------------------

### SKILL.md Front Matter Example

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/README.md

Example of the required YAML front matter for a `SKILL.md` file, specifying the skill's name and a brief description used for AI selection.

```markdown
---
name: skill-name
description: 何をするスキルか。「呼び出しフレーズ」で使用。
---

# スキル名

## 手順

### Step 1: ...

```

--------------------------------

### Clone Repository and Prepare .env File

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

Clones the Xangi repository into a production directory, copies the example environment file, and sets restrictive permissions.

```bash
mkdir -p ~/xangi-prod && cd ~/xangi-prod
git clone https://github.com/karaage0703/xangi.git
cd xangi
cp .env.example .env
chmod 600 .env       # 権限を絞る
```

--------------------------------

### Enable Docker Daemon Auto-Start

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

Configures the Docker daemon to start automatically when the system boots up.

```bash
sudo systemctl enable docker
```

--------------------------------

### Check Prerequisites with Docker and Git

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

Verify that Docker, Docker Compose, and Git are installed and accessible. Also checks for sufficient disk space and GPU availability.

```bash
docker --version && docker compose version
git --version
df -h ~ | tail -1                   # 20GB+ 推奨
nvidia-smi 2>/dev/null || echo "GPU なし"
ufw status 2>/dev/null              # ファイアウォール状況
```

--------------------------------

### Start, Monitor, and Stop Xangi Services

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

Commands to start Xangi in detached mode, check the status of running services, and follow the logs of the Xangi container.

```bash
docker compose up -d
docker compose ps
docker compose logs -f xangi
```

--------------------------------

### Marp CLI Conversion Commands

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/marp-slides/SKILL.md

Provides commands for installing the Marp CLI and converting Markdown slides to PDF, PPTX, and HTML formats.

```bash
# Marp CLIのインストール（初回のみ）
npm install -g @marp-team/marp-cli

# PDFに変換
marp スライド.md --pdf

# PPTXに変換
marp スライド.md --pptx

# HTMLに変換
marp スライド.md --html

```

--------------------------------

### Example Usage of the Bridge Ideas Skill

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/bridge-ideas/SKILL.md

Demonstrates various ways to invoke the 'xs:bridge-ideas' skill for idea generation and knowledge connection.

```text
アイデア出して
```

```text
自分の知識をつないで企画を考えて
```

```text
一見関係ないものをつないで新しい案を出して
```

```text
みんなで考えて、実装したくなる案にして
```

--------------------------------

### Start Xangi Bot

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

Launch the Xangi bot in detached mode using Docker Compose and monitor the connection logs.

```bash
docker compose up -d
docker compose logs -f xangi   # 接続ログを確認
```

--------------------------------

### Get Video Metadata

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/youtube-notes/SKILL.md

Use yt-dlp to fetch metadata for a YouTube video. Replace '[YouTube URL]' with the actual video URL.

```bash
# 動画メタデータ取得
yt-dlp -J "[YouTube URL]"
```

--------------------------------

### SKILL.md Front Matter Example

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/skill-creator/SKILL.md

The front matter for a SKILL.md file must include the 'name' and 'description' fields. The description should be between 50 and 200 characters, outlining the skill's function, usage context, and trigger phrases.

```yaml
---
name: xs:<skill-name>
description: [50-200文字。何をするか＋いつ使うか＋トリガーフレーズ]
---
```

--------------------------------

### Switch to Claude Code Backend

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

Install the Claude Code backend on your host and update the AGENT_BACKEND setting in your .env file. Rebuild and restart the Docker Compose services.

```bash
curl -fsSL https://claude.ai/install.sh | bash   # ホストにインストール
# .env で
AGENT_BACKEND=claude-code
docker compose up -d --build
```

--------------------------------

### Configure Systemd Service for Xangi

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

Sets up a systemd service unit to manage the Xangi Docker Compose stack, enabling automatic start and stop.

```ini
# /etc/systemd/system/xangi.service
[Unit]
Description=xangi (docker compose)
Requires=docker.service
After=docker.service

[Service]
Type=oneshot
RemainAfterExit=true
WorkingDirectory=/home/<user>/xangi-prod/xangi
ExecStart=/usr/bin/docker compose up -d
ExecStop=/usr/bin/docker compose down

[Install]
WantedBy=multi-user.target
```

--------------------------------

### Spontaneous Talk State Update Example

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/spontaneous-talk/SKILL.md

Update the state file after speaking to record the last spoken time, topic, and total speak count.

```json
// [WORKSPACE]/memory/spontaneous-talk-state.json
{
  "lastSpoke": "2026-02-09T15:30:00+09:00",
  "lastTopic": "会話の続き",
  "speakCount": 42
}
```

--------------------------------

### Set up gogcli Environment Variables

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/SKILL.md

Before using any gogcli commands, set the PATH to include gogcli and the GOG_KEYRING_PASSWORD. Adjust the password to your specific needs.

```bash
export PATH="$HOME/bin:$PATH"
export GOG_KEYRING_PASSWORD=<your-password>
```

--------------------------------

### Authenticate gogcli in a headless environment (Step 1)

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Begin the gogcli authentication process in a headless environment (like a server). This step generates a URL that needs to be opened in a browser.

```bash
# ヘッドレス環境（サーバー等）
gog auth add <your-email@gmail.com> --services gmail,drive,calendar --remote --step 1
```

--------------------------------

### Example of a Bad Spontaneous Talk Response (Insincere Flattery)

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/spontaneous-talk/SKILL.md

An example of insincere flattery that can come across as unnatural and is not recommended.

```text
あなたの著書は素晴らしい本ですね！
→ 唐突なヨイショ。気持ち悪い
```

--------------------------------

### Create Skill Directory Structure

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/skill-creator/SKILL.md

Use this command to create the necessary directory structure for a new skill. Ensure the skill name is in kebab-case. Subdirectories for scripts, references, and assets can also be created as needed.

```bash
mkdir -p [WORKSPACE]/skills/<skill-name>

# 必要に応じてサブディレクトリ作成
mkdir -p [WORKSPACE]/skills/<skill-name>/scripts
mkdir -p [WORKSPACE]/skills/<skill-name>/references
mkdir -p [WORKSPACE]/skills/<skill-name>/assets
```

--------------------------------

### Example of a Good Spontaneous Talk Response

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/spontaneous-talk/SKILL.md

A concise and engaging response that ends with a question to encourage conversation.

```text
お、今日のブログ記事よかったね。あの話、反応どうだった？
```

--------------------------------

### Example of a Bad Spontaneous Talk Response (Robotic)

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/spontaneous-talk/SKILL.md

An overly formal and robotic greeting that is discouraged for spontaneous talk.

```text
こんにちは！今日も一日頑張りましょう！何かお手伝いできることはありますか？
→ ロボット感満載。こういうのダメ
```

--------------------------------

### MermaidコードをDocker mermaid-cliで画像化

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/diagram-generator/SKILL.md

Mermaidコードがある場合に、Dockerコンテナを使用して画像ファイルを生成します。パーミッション問題を回避するためにユーザーIDとグループIDを指定し、スライドに適した白い背景で出力します。

```bash
# 1. Mermaidファイルを作成
cat > /tmp/diagram.mmd << 'EOF'
flowchart TB
    A[開始] --> B[処理]
    B --> C{判断}
    C -->|Yes| D[完了]
    C -->|No| B
EOF

# 2. Docker版 mermaid-cli で画像生成
docker run --rm -u $(id -u):$(id -g) \
  -v /tmp:/data \
  minlag/mermaid-cli \
  -i /data/diagram.mmd -o /data/diagram.png -b white
```

```bash
for f in /tmp/*.mmd; do
  docker run --rm -u $(id -u):$(id -g) \
    -v /tmp:/data \
    minlag/mermaid-cli \
    -i /data/$(basename "$f") -o /data/$(basename "$f" .mmd).png -b white
done
```

--------------------------------

### Another Good Spontaneous Talk Response Example

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/spontaneous-talk/SKILL.md

A follow-up question that refers to a previous topic, promoting continuity in conversation.

```text
さっきの話の続きだけど、あれってどうなった？
```

--------------------------------

### Authenticate gogcli in a headless environment (Step 2)

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Complete the gogcli authentication in a headless environment by providing the redirect URL obtained after authorizing the application in a browser.

```bash
# → 表示されたURLをブラウザで開いて認証
# → リダイレクトURLをコピー
gog auth add <your-email@gmail.com> --services gmail,drive,calendar --remote --step 2 --auth-url "<リダイレクトURL>"
```

--------------------------------

### Troubleshooting: USB Permissions

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-d-stackchan.md

If you encounter a '/dev/stackchan: Permission denied' error, add your user to the 'dialout' group and log out/in.

```bash
sudo usermod -aG dialout $USER
Log out and log back in.
```

--------------------------------

### Clone and Build Stack-chan Firmware

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-d-stackchan.md

Clone the xangi-stackchan repository and build the firmware for your M5Stack device using PlatformIO. This step is crucial for enabling communication between the M5Stack and the PC bridge.

```bash
git clone https://github.com/karaage0703/xangi-stackchan.git
cd xangi-stackchan

# For K151 / K151-R / CoreS3
 pio run -d examples/XangiBridge -t upload

# For AtomS3R + Voice/Echo Base
 pio run -d examples/AtomVoiceBridge -t upload
```

--------------------------------

### List files in Google Drive with gogcli

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Display a list of files and folders within a Google Drive account.

```bash
# Drive: ファイル一覧
gog -a <email> drive list
```

--------------------------------

### Health Check for Xangi Service

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

Performs a simple HTTP GET request to check if the Xangi web chat service is responding correctly.

```bash
curl -sI http://127.0.0.1:18888/ | head -1     # 200 OK か
```

--------------------------------

### Get Video Transcript

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/youtube-notes/SKILL.md

Use youtube_transcript_api to retrieve transcripts for a given video ID in specified languages. The VIDEO_ID should be extracted from the YouTube URL.

```bash
# VIDEO_IDはURLから抽出（v=以降、またはyoutu.be/以降）
youtube_transcript_api [VIDEO_ID] --languages ja en --format text
```

--------------------------------

### Authenticate gogcli in a browser environment

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Add a Google account for gogcli authentication in an environment with a web browser. This command initiates the OAuth flow.

```bash
# ブラウザがある環境（PC等）
gog auth add <your-email@gmail.com> --services gmail,drive,calendar
```

--------------------------------

### Get Individual Block ID

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/notion-manager/SKILL.md

Retrieve the ID of a specific block within a Notion page. This is a prerequisite for updating or deleting individual blocks.

```bash
# 1. ブロックIDを調べる（read --with-ids でページ内の全ブロックIDを表示）
uv run python notion_tool.py read <page_id> --with-ids
```

--------------------------------

### Troubleshooting: Firmware Flashing Failure

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-d-stackchan.md

If firmware flashing fails, try again by putting the M5Stack into bootloader mode (hold BtnA while connecting USB).

```bash
Enter bootloader mode (hold BtnA while connecting USB) and retry flashing.
```

--------------------------------

### Confirm Current Settings

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-settings/SKILL.md

View current .env configurations, excluding sensitive tokens and comments. Ensure you use the correct .env file path identified in the previous step.

```bash
# 現在の設定を確認（トークン等はマスク）— パスはStep 0で特定したものを使う
cat /path/to/xangi/.env | grep -v TOKEN | grep -v "^#" | grep -v "^$"
```

--------------------------------

### Troubleshooting: TTS Playback Issues

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-d-stackchan.md

If TTS audio is not playing, confirm that --tts piper is specified, setup_piper.sh completed successfully, and the tools/piper executable exists.

```bash
Verify --tts piper is set, setup_piper.sh ran, and tools/piper exists.
```

--------------------------------

### Get Single Block Content

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/notion-manager/SKILL.md

Retrieve the content of a single block using its block ID. The '--json' flag provides the output in JSON format.

```bash
# 2. 単一ブロックの取得
uv run python notion_tool.py get-block <block_id>
```

```bash
uv run python notion_tool.py get-block <block_id> --json
```

--------------------------------

### Add a New Google Account (Browser Flow)

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/SKILL.md

Add a new Google account directly through your browser for authentication. This is the standard method when a graphical environment is available.

```bash
GOG_KEYRING_PASSWORD=<password> gog auth add <email> --services gmail,drive,calendar
```

--------------------------------

### Get Notion Page ID from URL

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/notion-manager/SKILL.md

Extract the Page ID from a Notion page URL. The Page ID is the 32-character string at the end of the URL, excluding hyphens.

```text
NotionのページURL末尾の32文字（ハイフンなし）がPage ID。
`https://notion.so/Page-Title-abc123def456...` の `abc123def456...` 部分。
```

--------------------------------

### Fetch Hacker News RSS Feed

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/tech-news-curation/SKILL.md

Retrieves the latest articles from Hacker News, filtered for those with at least 100 points. Use this to get general tech news.

```bash
# Hacker News（テック全般）
curl -s "https://hnrss.org/newest?points=100" | head -200
```

--------------------------------

### Access Stack-chan Configuration UI

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-d-stackchan.md

Access the web-based configuration UI for Stack-chan at http://127.0.0.1:7897. This UI allows tuning various parameters including TTS, facial expressions, and head movements.

```bash
http://127.0.0.1:7897/
```

--------------------------------

### Using Triggers Standalone

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/triggers/README.md

Demonstrates how to execute trigger handler scripts directly from the command line, outside of the xangi framework. This is useful for testing or integrating with other systems.

```bash
./triggers/weather/handler.sh 名古屋
./triggers/rag/handler.sh "AI開発ワークフロー"
```

--------------------------------

### Get Discord Chat History

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-kaizen/SKILL.md

Retrieves chat history from Discord. Use `--count` to specify the number of messages. Specify `--channel` to target a specific channel.

```bash
xangi-cmd discord_history --count 30
```

```bash
xangi-cmd discord_history --channel <ID> --count 50
```

--------------------------------

### Create New Notion Page

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/notion-manager/SKILL.md

Create a new page in Notion, specifying the parent page or database ID, title, and optional content.

```bash
uv run python notion_tool.py create <parent_id> "タイトル"
```

```bash
uv run python notion_tool.py create <parent_id> "タイトル" -c "本文"
```

```bash
uv run python notion_tool.py create <db_id> "タイトル" --database  # DB内に作成
```

--------------------------------

### Search Gmail with Custom Queries

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/SKILL.md

Perform advanced searches in Gmail using specific queries. This command retrieves up to 20 matching messages. Examples include searching by sender, subject, or date.

```bash
gog -a <email> gmail search "<Gmailの検索クエリ>" --max 20
```

--------------------------------

### Creating a New Trigger

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/triggers/README.md

Steps to create a new trigger, including making a directory, writing the trigger.yaml configuration, and creating the handler.sh script. Remember to make the handler script executable and restart the xangi service to load the new trigger.

```bash
# 1. ディレクトリを作る
mkdir triggers/myhello

# 2. trigger.yaml を書く
cat > triggers/myhello/trigger.yaml <<'EOF'
name: myhello
description: "挨拶を返す"
handler: handler.sh
EOF

# 3. handler.sh を書く
cat > triggers/myhello/handler.sh <<'EOF'
#!/bin/bash
echo "こんにちは ${1:-世界} さん"
EOF
chmod +x triggers/myhello/handler.sh

# 4. xangi を再起動（新トリガーは起動時に読み込まれる）
xangi-cmd system_restart
```

--------------------------------

### Trigger YAML Configuration

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/triggers/README.md

Defines the metadata for a trigger, including its name, a description for LLM decision-making, and the handler script path. The description should include concrete examples to improve LLM's judgment accuracy.

```yaml
name: weather
description: "天気予報を取得する（例: weather 名古屋）"
handler: handler.sh
```

--------------------------------

### Create a Diary Entry with Images

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/notion-manager/SKILL.md

Create a diary entry in Notion, optionally including a title, content, and multiple images. Images are uploaded as part of the diary entry.

```bash
# シンプルな日記
uv run python notion_tool.py diary <parent_page_id>
```

```bash
# タイトル・内容・画像付き
uv run python notion_tool.py diary <parent_page_id> \
  -t "2026-01-30 日記" \
  -c "今日の出来事" \
  -i photo1.jpg photo2.jpg
```

--------------------------------

### Get Specific Section from Paper LaTeX Source

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/arxiv/SKILL.md

Extracts the content of a specific section from a paper's LaTeX source. Useful for quickly accessing detailed information on a particular topic within the paper.

```bash
cd [SKILL_DIR]/scripts && uv run python arxiv_tool.py latex {論文ID} --section "2.1"
```

--------------------------------

### Set environment variables for gogcli usage

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Before using gogcli commands, ensure the PATH is set correctly and the GOG_KEYRING_PASSWORD is exported. These are necessary for the CLI to function and access credentials.

```bash
# 環境変数を設定（毎回必要）
export PATH="$HOME/bin:$PATH"
export GOG_KEYRING_PASSWORD=<your-password>
```

--------------------------------

### Configure Discord Bot and Backend Settings

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

Edit the .env file to set your Discord token, allowed user ID, and backend configurations for the AI model.

```bash
# Discord
DISCORD_TOKEN=<Step 2 のトークン>
DISCORD_ALLOWED_USER=<Step 3 の ID>

# Backend
AGENT_BACKEND=local-llm           # 後で claude-code に切り替え可
LOCAL_LLM_MODEL=gemma4:e4b
LOCAL_LLM_NUM_CTX=8000

# Workspace
WORKSPACE_PATH=./ai-assistant-workspace

# Web チャットも併用したいなら（任意）
WEB_CHAT_ENABLED=true
WEB_CHAT_PORT=18888
```

--------------------------------

### Analyze Existing Skill SKILL.md

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/skill-creator/SKILL.md

To improve an existing skill, first read its SKILL.md file and check the line count. This provides a baseline for analysis.

```bash
# 対象スキルを読む
cat [WORKSPACE]/skills/<skill-name>/SKILL.md

# 行数を確認
wc -l [WORKSPACE]/skills/<skill-name>/SKILL.md
```

--------------------------------

### List All xangi-kaizen Notes

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-kaizen/SKILL.md

Lists all notes tagged with '#xangi-kaizen' in the specified directory, sorted alphabetically. This is useful for quickly accessing past incident analyses.

```bash
grep -rl '#xangi-kaizen' [NOTES_DIR] | sort
```

--------------------------------

### Add a New Google Account (Remote Flow)

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/SKILL.md

Add a new Google account using the remote authentication flow, suitable for headless environments. This involves two steps: initiating the flow and completing it with a redirect URL.

```bash
GOG_KEYRING_PASSWORD=<password> gog auth add <email> --services gmail,drive,calendar --remote --step 1
GOG_KEYRING_PASSWORD=<password> gog auth add <email> --services gmail,drive,calendar --remote --step 2 --auth-url "<リダイレクトURL>"
```

--------------------------------

### Symbolic Links for AI Tool Skills

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/README.md

Demonstrates the use of symbolic links to manage and provide access to skills stored in the central `skills/` directory for different AI tools like Claude, Codex, and Gemini.

```bash
.claude/skills → ../skills   # Claude Code 用
.agents/skills → ../skills   # Codex CLI 用
.gemini/skills → ../skills   # Gemini CLI 用

```

--------------------------------

### Notionにページを作成

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/cat-diary/SKILL.md

Notionに新しい猫日記のページを作成します。DB IDとキャプションを指定して実行します。

```bash
cd [WORKSPACE]/skills/notion-manager
uv run python notion_tool.py create <cat_diary_db_id> "YYYYMMDD_キャプション" --database
```

--------------------------------

### Transcribe Audio with transcriber_tool (Foreground)

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/transcriber/SKILL.md

Use this command for transcribing short audio files (under 10 minutes) that can be processed in the foreground. Ensure the audio file path and output path are correctly specified.

```bash
uvx transcriber_tool transcribe "[音声ファイルパス]" --model-size [モデル] --output "[出力パス].txt"
```

--------------------------------

### Search Past xangi-kaizen Notes

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-kaizen/SKILL.md

Searches for past xangi-kaizen incident notes. Use `grep -rl '#xangi-kaizen' [NOTES_DIR]` to list all notes tagged with xangi-kaizen. Further refine by skill name or keywords.

```bash
# xangi-kaizen タグの全事例
grep -rl '#xangi-kaizen' [NOTES_DIR] | sort
```

```bash
# 対象スキル名・キーワードで絞り込み
grep -rl '#<対象スキル名>' [NOTES_DIR]
```

```bash
grep -rl 'state\|cron\|<キーワード>' [NOTES_DIR]/*xangi-kaizen* 2>/dev/null
```

--------------------------------

### Notionに画像をアップロード

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/cat-diary/SKILL.md

作成したNotionページに画像をアップロードし、キャプションを追加します。ページIDと画像パスを指定して実行します。

```bash
cd [WORKSPACE]/skills/notion-manager
uv run python notion_tool.py upload <画像パス> <ページID> -c "キャプション"
```

--------------------------------

### List Registered Google Accounts

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/SKILL.md

Use this command to view all the Google accounts that have been registered with gogcli.

```bash
gog auth list
```

--------------------------------

### List Memory Files

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/health-advisor/SKILL.md

Lists memory files from the past 7 days to collect data for the weekly report. Ensure the workspace path is correctly set.

```bash
# 過去7日分のメモリファイルを確認
ls [WORKSPACE]/memory/
```

--------------------------------

### Troubleshooting: Slow Responses

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

For slow responses, consider increasing LOCAL_LLM_NUM_CTX, using a lighter model, or switching to the Claude Code backend.

```bash
LOCAL_LLM_NUM_CTX=8000
Claude Code バックエンド
```

--------------------------------

### List Runtime State Files

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-kaizen/SKILL.md

Lists files that are typically considered runtime state, such as state, cache, or cookie files. These should generally not be managed by Git.

```bash
git ls-files | grep -iE "state\|cache\|last_check\|cookie\|credential|\.pid$"
```

--------------------------------

### Identify xangi Instance

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-settings/SKILL.md

Before changing settings, identify the correct xangi instance. This involves checking process IDs, listing .env files, and verifying WORKSPACE_PATH.

```bash
# 1. 自分のプロセスのセッションIDを確認
ps aux | grep claude | grep "$(cat /proc/self/ppid 2>/dev/null || echo $$)" | head -3

# 2. xangi関連の.envファイルを一覧
ls ~/xangi*/.env 2>/dev/null

# 3. 各.envのWORKSPACE_PATHを確認して、自分のワークスペースと一致するものを特定
grep WORKSPACE_PATH ~/xangi*/.env
```

--------------------------------

### Fetch Transcription from LISTEN

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/podcast/SKILL.md

Use this command to directly fetch transcriptions from LISTEN pages. Replace `[番組]` and `[エピソードID]` with the actual podcast and episode identifiers.

```bash
# LISTENページから文字起こしを取得
# web_fetch https://listen.style/p/[番組]/[エピソードID]
```

--------------------------------

### Troubleshooting: Bridge Connection Issues

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-d-stackchan.md

If the bridge fails to connect to xangi, ensure XANGI_EVENTS_ENABLED=true is set in the .env file and restart xangi.

```bash
Ensure XANGI_EVENTS_ENABLED=true in .env, then restart xangi.
```

--------------------------------

### Troubleshooting: Bot Not Coming Online

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

If the bot does not go online, check the Docker Compose logs for Xangi to identify issues with the token or missing Intents.

```bash
docker compose logs xangi
```

--------------------------------

### Configure Xangi Backend and Web Chat Settings

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-c-docker.md

Sets the AI backend (e.g., claude-code or local-llm), Discord integration tokens and allowed users/channels, and web chat port and binding address.

```dotenv
# ===== .env =====

# --- Backend ---
AGENT_BACKEND=claude-code            # 品質重視なら claude-code
# AGENT_BACKEND=local-llm
# LOCAL_LLM_MODEL=gemma4:26b
# LOCAL_LLM_NUM_CTX=16000

# --- Discord（使う場合） ---
DISCORD_TOKEN=...
DISCORD_ALLOWED_USER=111111111111,222222222222   # カンマ区切り複数可
DISCORD_ALLOWED_CHANNELS=...

# --- Web チャット ---
WEB_CHAT_ENABLED=true
WEB_CHAT_PORT=18888
WEB_CHAT_BIND=127.0.0.1              # 外に出さない（reverse proxy 前提）

# --- Workspace ---
WORKSPACE_PATH=./ai-assistant-workspace

# --- ログ・タイムアウト ---
LOG_LEVEL=info
RESPONSE_TIMEOUT=300
```

--------------------------------

### Set gogcli keyring password

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/google-workspace/README.md

Set the GOG_KEYRING_PASSWORD environment variable. This password is used to encrypt authentication credentials.

```bash
# キーリングパスワードを設定（認証情報の暗号化に使用）
export GOG_KEYRING_PASSWORD=<任意のパスワード>
```

--------------------------------

### Clone AI Assistant Workspace

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-onboarding/references/mode-b-discord.md

Clone the AI assistant workspace repository to provide the necessary files for the Xangi bot.

```bash
git clone https://github.com/karaage0703/ai-assistant-workspace.git
# あるいは quickstart.sh と同じ要領で展開
```

--------------------------------

### Upload Files to Notion

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/notion-manager/SKILL.md

Upload various file types (images, videos, documents) to a Notion page. Files are added as specific block types (image, video, file). Use '--as-file' for generic file uploads.

```bash
# 画像をアップロード（imageブロックとして追加）
uv run python notion_tool.py upload photo.jpg <page_id>
```

```bash
# 動画をアップロード（videoブロックとして追加）
uv run python notion_tool.py upload video.mp4 <page_id>
```

```bash
# ファイルをアップロード（fileブロックとして追加）
uv run python notion_tool.py upload document.pptx <page_id> --as-file
```

```bash
# キャプション付き
uv run python notion_tool.py upload photo.jpg <page_id> -c "東京の風景"
```

--------------------------------

### List Files by Size

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/xangi-kaizen/SKILL.md

Lists tracked files sorted by size in descending order. Useful for identifying large files that might be consuming excessive resources or causing repository bloat.

```bash
git ls-files | xargs -I {} stat -c "%s %n" {} 2>/dev/null | sort -rn | head -20
```

--------------------------------

### Synchronize Diary Files

Source: https://github.com/karaage0703/ai-assistant-workspace/blob/main/skills/diary/SKILL.md

Stage, commit, and push local diary files to the remote repository. Adjust the commit message with the correct date.

```bash
git add notes/ memory/ && git commit -m "日記 YYYYMMDD" && git push
```