The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module in scripts/aria2_acquire.py and scripts/verify_media.py to interact with aria2c and ffprobe. These calls are implemented using list-based arguments without the shell=True parameter, which effectively mitigates the risk of shell injection from untrusted inputs like media titles or magnet URIs.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from external media sources and search results. It implements appropriate safeguards to prevent this data from influencing the agent's logic or system security. Evidence chain: (1) Ingestion points: External magnet URIs, torrent files, and search result JSON in scripts/qbittorrent_search.py; (2) Boundary markers: The skill enforces job isolation using specific probe directories and unique job IDs; (3) Capability inventory: Subprocess calls are restricted to the aria2c and ffprobe binaries for download and verification tasks; (4) Sanitization: Input data is treated as literal strings within subprocess argument lists.
[SAFE]: The skill handles configuration and credentials securely, explicitly advising users to store sensitive information like qBittorrent passwords in environment variables or the operating system's credential store rather than in code or configuration files. No evidence of data exfiltration, obfuscation, or persistence mechanisms was found.

lov-media-fetch