edgespeak-yt-download
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (yt-dlp, ffmpeg, ffprobe) that incorporate user-supplied input such as video URLs, output paths, and format strings. If these inputs are not strictly validated by the agent platform, it could lead to command injection.
- [INDIRECT_PROMPT_INJECTION]: The skill fetches external, untrusted data including YouTube metadata (.info.json) and subtitle files. This content is passed to other agent components, creating an attack surface where malicious instructions embedded in a video's metadata could influence subsequent agent behavior.
- Ingestion points: Video metadata and caption files downloaded from YouTube (referenced in SKILL.md).
- Boundary markers: Absent; no explicit delimiters or instructions are provided to the agent to treat this external content as non-executable text.
- Capability inventory: The skill utilizing subprocess calls to interact with media processing CLI tools.
- Sanitization: Absent; the skill does not specify any programmatic sanitization or filtering of the metadata before it is passed to downstream skills.
- [EXTERNAL_DOWNLOADS]: The skill performs network requests to fetch media and metadata from YouTube, a well-known service. It also recommends updating yt-dlp from official sources. These are documented as standard functional requirements for the skill's purpose.
Audit Metadata