The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation details how to download and install the yt-dlp binary and its components from official sources.
Evidence: wiki/Installation.md provides curl and wget commands targeting github.com/yt-dlp/yt-dlp.
Evidence: wiki/EJS.md describes the --remote-components flag for fetching challenge solvers from github.com/yt-dlp/ejs.
[COMMAND_EXECUTION]: The skill includes extensive documentation for executing CLI commands using the yt-dlp tool to manage video and audio downloads.
[DATA_EXFILTRATION]: The documentation describes the use of authentication features that access sensitive local data, such as browser cookies and the user's .netrc file, which is the intended primary purpose of the tool for accessing restricted content.
Evidence: README.md and wiki/FAQ.md cover the --cookies-from-browser and --netrc options.
[INDIRECT_PROMPT_INJECTION]: The skill acts as an interface for a tool that ingests untrusted data from the web (video titles, descriptions, and comments). This creates a known attack surface for indirect prompt injection if the agent processes this output without sufficient isolation.
Ingestion points: Video metadata and external URLs processed by the extractor (SKILL.md, README.md).
Boundary markers: None explicitly defined in the documentation.
Capability inventory: The skill uses yt-dlp, which possesses network capabilities, filesystem write access, and access to local credentials.
Sanitization: Not explicitly implemented in the skill; depends on the calling agent's safety protocols.

yt-dlp