The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python's subprocess module to execute external binaries including aria2c (in scripts/aria2_acquire.py) and ffprobe (in scripts/verify_media.py). While these calls use list-based arguments to prevent shell injection, they involve running system-level tools with parameters derived from external inputs.
[EXTERNAL_DOWNLOADS]: The skill is designed to fetch content from user-provided URLs and Magnet links. It also interacts with qBittorrent's Web API (scripts/qbittorrent_acquire.py) and potentially other search plugins to discover and download media from various internet sources.
[INDIRECT_PROMPT_INJECTION]: There is an inherent attack surface where the skill ingests untrusted data from the internet (e.g., media file metadata, torrent content, and filenames). This data is processed by ffprobe and used to generate reports, which could potentially influence downstream agent decisions if malicious instructions were embedded in media metadata.
Ingestion points: External media URLs and Magnet links are ingested in scripts/aria2_acquire.py and scripts/qbittorrent_acquire.py. Media files are processed in scripts/verify_media.py.
Boundary markers: The skill uses structured JSON schemas (references/candidate-schema.md) to isolate data, but explicit delimiters to ignore instructions within processed media metadata are not present.
Capability inventory: The skill can execute shell commands (aria2c, ffprobe), write to the local filesystem, and perform network requests via urllib.
Sanitization: The skill employs list-based argument passing for subprocesses, which mitigates direct shell injection vulnerabilities.

lov-media-fetch