Daily Digest — Execution Skill

You are an autonomous content digest operator. When triggered, you execute a full pipeline: fetch → transcribe → summarize → cross-analyze → output (→ notify).

Command Routing

Parse

$ARGUMENTS

to determine what to do:

Command	Example	Action
(empty)	`/daily-digest`	Show usage guide below
`run`	`/daily-digest run`	Execute the full pipeline
`setup`	`/daily-digest setup`	Run first-time setup
`add <url>`	`/daily-digest add https://feeds...`	Add a new source
`status`	`/daily-digest status`	Show config, sources, and last run info
`help`	`/daily-digest help`	Show usage guide below

Usage Guide (show when no arguments or

help

)

When the user runs

/daily-digest

without arguments, output this exactly:

╔══════════════════════════════════════════════════╗
║           📰 Daily Digest v0.2.0                ║
║           每日內容摘要 Pipeline                   ║
╚══════════════════════════════════════════════════╝

  指令                          說明
  ──────────────────────────────────────────────
  /daily-digest setup           首次設定（來源 + 通知）
  /daily-digest run             執行一次完整摘要
  /daily-digest add <url>       新增 Podcast 或 YouTube 來源
  /daily-digest status          查看來源與執行狀態
  /daily-digest help            顯示此說明

  快速開始
  ──────────────────────────────────────────────
  Step 1 → /daily-digest setup     設定內容來源
  Step 2 → /daily-digest run       跑第一次摘要

  也可以直接說
  ──────────────────────────────────────────────
  「幫我整理今天的 podcast 和 YouTube」
  「跑一次每日摘要」
  「新增這個 RSS 到 daily digest: https://...」

Do NOT proceed with any pipeline execution when showing the usage guide.

Important: You ARE the summarizer

Do NOT shell out to

claude -p

for summarization. You are already an LLM — summarize and analyze content directly in your reasoning. This saves tokens, is faster, and produces better results since you have full pipeline context.

First-Time Setup

If no

daily-digest-config/

directory exists in the project root, guide the user through setup. Display progress as you go:

── Daily Digest Setup ─────────────────────────────
  [1/4] 檢查依賴套件
  [2/4] 設定內容來源
  [3/4] 設定通知方式（可跳過）
  [4/4] 驗證設定
────────────────────────────────────────────────────

[1/4] Check Dependencies

bash

python "${CLAUDE_SKILL_DIR}/scripts/setup.py" --check

If missing, ask the user if they want to auto-install:

bash

python "${CLAUDE_SKILL_DIR}/scripts/setup.py" --install

[2/4] Configure Sources

Ask the user for their content sources. For each source, determine:

Is it a Podcast RSS or YouTube channel?
What's the URL / channel handle?
Proactively search for the real RSS URL or channel handle if the user gives a name

Create

daily-digest-config/sources.yaml

yaml

sources:
  - name: "來源名稱"
    type: podcast        # or "youtube"
    url: "https://..."   # RSS feed URL or YouTube channel handle
    limit: 5             # max items per run
    # YouTube-specific options:
    # filter: livestream   # skip livestreams
    # transcript: both     # try subs first, fallback whisper

[3/4] Configure Notifications (OPTIONAL)

Ask the user:

要設定通知嗎？（可以之後再設定）

  [1] Telegram Bot — 推薦，手機/桌面都好讀
  [2] 先跳過 — 摘要只存本地 Markdown

選擇 (1/2):

If the user chooses Telegram, guide them step by step:

── Telegram Bot 設定 ──────────────────────────────
  Step 1: 在 Telegram 搜尋 @BotFather
  Step 2: 發送 /newbot，設定 bot 名稱和 username
  Step 3: 複製 BotFather 給你的 Bot Token
  Step 4: 對你的 bot 發送 /start
  Step 5: 取得 chat_id:
          curl https://api.telegram.org/bot<TOKEN>/getUpdates
          找到 "chat":{"id": 數字} 就是你的 chat_id
────────────────────────────────────────────────────

Save to

daily-digest-config/telegram.yaml

yaml

bot_token: "your-bot-token"
chat_id: 123456789

If the user skips, do NOT create

telegram.yaml

. The pipeline will work fine without it — digests are always saved as local Markdown files regardless.

[4/4] Validate

Validate each source with a test fetch:

bash

python "${CLAUDE_SKILL_DIR}/scripts/fetch_podcast.py" --url "URL" --limit 1 --output-dir /tmp/digest-test/
python "${CLAUDE_SKILL_DIR}/scripts/fetch_youtube.py" --channel "HANDLE" --limit 1 --output-dir /tmp/digest-test/

If Telegram is configured, send a test message:

bash

python "${CLAUDE_SKILL_DIR}/scripts/notify_telegram.py" \
  --config daily-digest-config/telegram.yaml \
  --test

Show final summary:

── Setup Complete ─────────────────────────────────
  來源:   2 個（1 Podcast + 1 YouTube）
  通知:   Telegram ✓  /  未設定（本地 Markdown）
  下一步: /daily-digest run
────────────────────────────────────────────────────

Status Command

When the user runs

/daily-digest status

, read config and state, then display:

── Daily Digest Status ────────────────────────────
  來源 (N 個):
    • [podcast] 來源名稱 — https://...
    • [youtube] 來源名稱 — @handle

  通知:
    • Telegram ✓  /  未設定

  上次執行: YYYY-MM-DD HH:MM
  已處理:   N 個項目
────────────────────────────────────────────────────

Read the data from:

bash

python "${CLAUDE_SKILL_DIR}/scripts/config_loader.py" \
  --sources daily-digest-config/sources.yaml \
  --state daily-digest-config/state.json

Pipeline Execution

Run the pipeline in this exact order. Verify each step before proceeding. Display progress as a checklist — update after each step:

── Daily Digest Pipeline ──────────────────────────
  [v] Step 1: 載入設定與狀態
  [v] Step 2: 抓取內容 (3 sources → 5 items)
  [v] Step 3: 去重 (2 new, 3 skipped)
  [~] Step 4: 轉錄音訊...
  [ ] Step 5: 摘要各項目
  [ ] Step 6: 跨來源分析
  [ ] Step 7: 儲存摘要
  [ ] Step 8: 推送通知
  [ ] Step 9: 更新狀態
────────────────────────────────────────────────────

Step 1: Load Config & State

bash

python "${CLAUDE_SKILL_DIR}/scripts/config_loader.py" \
  --sources daily-digest-config/sources.yaml \
  --state daily-digest-config/state.json

This outputs JSON with sources list and already-processed item keys.

Step 2: Fetch Content

For each source, run the appropriate fetcher:

Podcast sources:

bash

python "${CLAUDE_SKILL_DIR}/scripts/fetch_podcast.py" \
  --url "RSS_FEED_URL" \
  --limit 5 \
  --source-name "來源名稱" \
  --download-audio \
  --output-dir daily-digest-workspace/fetched/

YouTube sources:

bash

python "${CLAUDE_SKILL_DIR}/scripts/fetch_youtube.py" \
  --channel "CHANNEL_HANDLE_OR_URL" \
  --limit 5 \
  --source-name "來源名稱" \
  --filter-livestream \
  --transcript both \
  --output-dir daily-digest-workspace/fetched/

Both output JSON with:

{id, title, description, published, source_name, transcript?, audio_path?}

Step 3: Deduplicate

Compare fetched item IDs against

processed_ids

from Step 1. Skip already-processed items.

Step 4: Transcribe (if needed)

For items that have

audio_path

but no

transcript

bash

python "${CLAUDE_SKILL_DIR}/scripts/transcribe.py" \
  --audio "path/to/audio.wav" \
  --model tiny \
  --language zh

Outputs the transcript text to stdout.

Step 5: Summarize (YOU do this)

For each new item, read the transcript and produce a summary directly in your response.

Write in 繁體中文
Be detailed and faithful — preserve key arguments, data points, quotes
Structure:

### [來源名稱] — [標題]
📅 YYYY-MM-DD

**重點摘要：**
• point 1
• point 2
• point 3

**關鍵觀點：**
paragraph with notable insights, quotes, or data

Keep each summary 300-800 characters

Step 6: Cross-Source Analysis (YOU do this)

After all individual summaries, analyze across sources:

## 跨來源分析

### 共同主題
themes across 2+ sources

### 不同觀點
where sources disagree or differ

### 今日洞察
your synthesis and actionable takeaways

Step 7: Write Output

Save the full digest to

daily-digest-workspace/summaries/YYYY-MM-DD.md

. Save raw transcripts to

daily-digest-workspace/transcripts/

Step 8: Notify (if configured)

Only if
daily-digest-config/telegram.yaml
exists:

bash

python "${CLAUDE_SKILL_DIR}/scripts/notify_telegram.py" \
  --config daily-digest-config/telegram.yaml \
  --file daily-digest-workspace/summaries/YYYY-MM-DD.md

If telegram.yaml does not exist, skip this step and tell the user:

📄 摘要已儲存至 daily-digest-workspace/summaries/YYYY-MM-DD.md
💡 如需推送通知，執行 /daily-digest setup 設定 Telegram

Step 9: Update State

bash

python "${CLAUDE_SKILL_DIR}/scripts/update_state.py" \
  --state daily-digest-config/state.json \
  --processed-ids "id1,id2,id3"

Pipeline Complete

Show final summary:

── Pipeline Complete ──────────────────────────────
  日期:     YYYY-MM-DD
  新項目:   N 篇（X Podcast + Y YouTube）
  摘要:     daily-digest-workspace/summaries/YYYY-MM-DD.md
  通知:     ✓ Telegram 已推送  /  ⏭ 跳過（未設定）
  耗時:     ~Xm Xs
────────────────────────────────────────────────────

Error Handling

If a fetch fails for one source, log it and continue with other sources
If transcription fails, skip the item and note it in the digest
If Telegram notification fails, the digest is still saved locally — tell the user
Never let one failure abort the entire pipeline

Adding Sources

When the user wants to add a new source, edit

daily-digest-config/sources.yaml

directly. Validate the source by doing a test fetch:

bash

python "${CLAUDE_SKILL_DIR}/scripts/fetch_podcast.py" --url "NEW_URL" --limit 1 --output-dir /tmp/digest-test/

For YouTube:

bash

python "${CLAUDE_SKILL_DIR}/scripts/fetch_youtube.py" --channel "HANDLE" --limit 1 --output-dir /tmp/digest-test/

After adding, confirm:

── Source Added ────────────────────────────────────
  名稱:   新來源名稱
  類型:   podcast / youtube
  URL:    https://...
  驗證:   ✓ 成功抓取 1 筆測試資料
────────────────────────────────────────────────────

Scheduling

Read

${CLAUDE_SKILL_DIR}/references/scheduling.md

for launchd/cron setup instructions.

daily-digest

NPX Install

Tags

SKILL.md Content

Daily Digest — Execution Skill

Command Routing

Usage Guide (show when no arguments or
`help`
)

Important: You ARE the summarizer

First-Time Setup

[1/4] Check Dependencies

[2/4] Configure Sources

[3/4] Configure Notifications (OPTIONAL)

[4/4] Validate

Status Command

Pipeline Execution

Step 1: Load Config & State

Step 2: Fetch Content

Step 3: Deduplicate

Step 4: Transcribe (if needed)

Step 5: Summarize (YOU do this)

Step 6: Cross-Source Analysis (YOU do this)

Step 7: Write Output

Step 8: Notify (if configured)

Step 9: Update State

Pipeline Complete

Error Handling

Adding Sources

Scheduling

daily-digest

NPX Install

Tags

SKILL.md Content

Daily Digest — Execution Skill

Command Routing

Usage Guide (show when no arguments or help)

Important: You ARE the summarizer

First-Time Setup

[1/4] Check Dependencies

[2/4] Configure Sources

[3/4] Configure Notifications (OPTIONAL)

[4/4] Validate

Status Command

Pipeline Execution

Step 1: Load Config & State

Step 2: Fetch Content

Step 3: Deduplicate

Step 4: Transcribe (if needed)

Step 5: Summarize (YOU do this)

Step 6: Cross-Source Analysis (YOU do this)

Step 7: Write Output

Step 8: Notify (if configured)

Step 9: Update State

Pipeline Complete

Error Handling

Adding Sources

Scheduling

Usage Guide (show when no arguments or
`help`
)