The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run() in scripts/download.py to invoke the yt-dlp utility. The command and its arguments are passed as a list, which prevents shell injection vulnerabilities.
[CREDENTIALS_UNSAFE]: The documentation in SKILL.md describes using the -u and -p flags to provide a username and password to yt-dlp. Hardcoding or passing credentials in plaintext via command-line arguments is an insecure practice as they can be captured in process lists or system logs.
[DATA_EXFILTRATION]: The skill documents and enables the use of the --cookies-from-browser flag. This feature allows the agent to access and extract sensitive session cookies from the user's web browsers, creating a risk for unauthorized data access or session hijacking if misused.
[INDIRECT_PROMPT_INJECTION]: The skill fetches data from arbitrary external URLs across thousands of websites. Malicious content or metadata (e.g., video titles or descriptions) hosted on these platforms could contain instructions designed to manipulate the agent's behavior.
Ingestion points: The url argument in scripts/download.py triggers the retrieval of external media data.
Boundary markers: The script does not implement specific delimiters or instructions to ignore content within the downloaded metadata.
Capability inventory: The skill has the ability to execute shell commands and perform file operations.
Sanitization: Filename sanitization is performed by yt-dlp, but no sanitization is applied to the metadata presented to the agent.
[EXTERNAL_DOWNLOADS]: The skill is designed to download media from a wide range of external sources and references the yt-dlp -U command for updating the downloader binary from remote repositories.

media-downloader