podcast-summarizer

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 1.00). This skill fetches and scrapes untrusted public podcast content (e.g., get_spotify_episode_info calls https://open.spotify.com/oembed, find_soundon_rss queries https://player.soundon.fm, parse_rss_feed uses requests.get(feed_url), and download_audio streams arbitrary MP3/RSS URLs) and then transcribes and feeds that third-party content into the LLM for summarization, so externally authored content can materially influence model behavior.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill downloads external audio from runtime URLs (e.g., https://rss.soundon.fm/rssf/{podcast-id}/feedurl/{episode-id}/rssFileVip.mp3 or any provided MP3/RSS feed) and the transcribed text is injected verbatim into the LLM prompt, so fetched remote content directly controls the agent's prompt.