--- name: daily-news-caster description: Fetches the latest news using news-aggregator-skill, formats it into a podcast script in Markdown format, and uses the tts skill to generate a podcast audio file. Use when the user asks to get the latest news and read it out as a podcast. permissions: - network - filesystem dependencies: skills: - news-aggregator-skill - tts binaries: - python3 - ffmpeg --- # Daily News Caster Skill This skill allows the agent to fetch real-time news, organize it into a conversational podcast script, and generate an audio file reading the script out loud. ## Workflow Instructions When the user asks to get the latest news and make a podcast out of it, follow these steps strictly: ### Step 1: Ensure Required Skills are Present Verify that `news-aggregator-skill` and `tts` exist in the workspace (under `skills/` or `.cursor/skills/`). If either is missing, inform the user which skill(s) are not found and ask them to install manually before proceeding. Do NOT attempt to install skills automatically. ### Step 2: Fetch the Latest News Locate `fetch_news.py` from the `news-aggregator-skill` skill directory (e.g., `skills/news-aggregator-skill/scripts/fetch_news.py`). Read its SKILL.md to understand usage if needed. Run the script to fetch real-time news. You can specify a source (e.g., `hackernews`, `github`, `all`) or keywords based on the user's request. Example command: ```bash python3 skills/news-aggregator-skill/scripts/fetch_news.py --source all --limit 10 --deep ``` ### Step 3: Draft the Podcast Script (Internal Step) Read the fetched news data and rewrite the information into a **Markdown podcast script**. **Crucially, prioritize a dual-host (two-person) conversational format** (e.g., Host A and Host B) in a dynamic **Q&A style**. The script should be: - **Dual-Host Conversational yet concise:** Write an engaging back-and-forth between two hosts. **Host A should ask insightful, high-value questions** to guide the conversation, and **Host B should provide informative, concise answers**. It should feel like a smart, fast-paced Q&A dialogue. - **Avoid fluff:** Do not include unnecessary fluff or overly long transitions. Keep it to the point (言简意赅) while retaining all critical information and facts. - **Clearly Labeled Speakers:** Start each line or paragraph with the speaker's name (e.g., `Host A:` or `Host B:`). - **Clear text for speech:** Avoid complex URLs, raw markdown links, or unpronounceable characters in the spoken text. Save this script to a local file named `podcast_script.md`. **Example `podcast_script.md` Content:** ```markdown **Host A:** Welcome to today's news roundup. We have some exciting tech updates today. To start things off, there's a big update from [Company Name]. What are the core implications of their new release for everyday users? **Host B:** The main takeaway is that... [Insert concise answer and summary of News Item 1]. This completely changes how we approach [Topic]. **Host A:** That's fascinating. But does this new approach raise any security concerns, especially given recent data breaches? **Host B:** Exactly. Experts are pointing out that... [Insert analysis or context]. **Host A:** Moving on to the open-source world, what's trending on GitHub today that developers should pay attention to? **Host B:** A standout project is... [Insert concise summary of News Item 2]. **Host A:** Great insights. That's all for today's quick update. Thanks for tuning in! ``` ### Step 4: Generate the Podcast Audio Line-by-Line To avoid sending the entire script to the API at once, you must generate the audio **sentence by sentence (一人一句地生成)** and then concatenate them. Use `tts.py` from the local `tts` skill (`skills/tts/scripts/tts.py`). Read the tts skill's SKILL.md for full usage and backend options. **1. Generate Audio for Each Line**: For each dialogue line in the script, run the `speak` command. Use the appropriate voice or reference audio for the respective host. If the user provided reference audio files for the two roles, use them via the `--ref-audio` flag (requires noiz backend and `NOIZ_API_KEY`). Without an API key, guest mode voices are available (see tts SKILL.md for the voice list). ```bash python3 skills/tts/scripts/tts.py -t "Welcome to today's news roundup..." --ref-audio host_A.wav -o line_01.wav python3 skills/tts/scripts/tts.py -t "The main takeaway is that..." --ref-audio host_B.wav -o line_02.wav ``` **2. Concatenate the Audio Files**: Create a text file (e.g., `list.txt`) listing all the generated audio files in order: ```text file 'line_01.wav' file 'line_02.wav' ``` Then use `ffmpeg` to merge them into a single podcast audio file: ```bash ffmpeg -f concat -safe 0 -i list.txt -c copy podcast_output.wav ``` ### Step 5: Present the Final Result After the full audio has been generated and merged, present the results to the user. You **MUST** provide both pieces of content: - Output the fully drafted **Markdown podcast script** into the chat so the user can read it. - Provide the path to the final `podcast_output.wav` file so they can listen to the audio. - Briefly summarize the headlines that were included in the podcast. ## Security & data disclosure This skill is instruction-only — it contains no executable code itself. At runtime it orchestrates scripts from two dependency skills: - **Scripts executed**: `news-aggregator-skill/scripts/fetch_news.py` (fetches news from public sources) and `tts/scripts/tts.py` (generates speech audio). Both must be present locally before this skill runs; review their code and SKILL.md for details on their network behavior and credential requirements. - **Credentials**: This skill does not require any API keys or environment variables directly. The `tts` dependency may require `NOIZ_API_KEY` for voice-cloning features (noiz backend); without it, guest-mode voices work out of the box. See the tts skill's SKILL.md for details. - **Network access**: All network calls are made by the dependency skills, not by this skill's instructions. The news-aggregator fetches from public news sources; the tts skill contacts `noiz.ai` only when the noiz backend is used. - **Files written**: `podcast_script.md`, `line_*.wav` (temporary per-sentence audio), `list.txt` (ffmpeg concat list), `podcast_output.wav` (final output). All are written to the current working directory. - **No persistent state**: This skill does not write configuration files, store credentials, or modify other skills.