The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes several system tools and package managers via subprocess calls.
scripts/setup.py executes pip install to manage the skill's Python dependencies.
scripts/fetch_podcast.py uses curl to download audio files from RSS feed URLs.
scripts/fetch_youtube.py utilizes yt-dlp to query video metadata and download subtitles or audio streams.
[EXTERNAL_DOWNLOADS]: The skill fetches content and software components from remote sources.
Media content (audio and text) is downloaded from user-configured RSS feeds and YouTube channels.
Python libraries are installed from the Python Package Index (PyPI) during the setup process.
The transcribe.py script triggers the download of Whisper models for speech-to-text processing.
[DATA_EXFILTRATION]: The skill transmits data to an external service as part of its primary function.
scripts/notify_telegram.py sends generated summaries to the Telegram Bot API using credentials (token and chat ID) provided by the user during configuration.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes untrusted data from external sources.
Ingestion points: Untrusted content enters the agent context via RSS feed summaries in fetch_podcast.py and YouTube transcripts (subtitles/audio) in fetch_youtube.py.
Boundary markers: The instructions in SKILL.md that direct the agent to summarize content do not employ explicit delimiters or 'ignore' instructions to isolate the source text from the agent's core logic.
Capability inventory: The skill environment possesses capabilities for file system writes, network requests, and system command execution (curl, yt-dlp, pip).
Sanitization: scripts/fetch_youtube.py includes a parse_vtt function that performs basic sanitization by stripping HTML tags from subtitle data.

daily-digest