The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a Python helper script, scripts/advproto.py, that executes the system's ffmpeg binary to perform streaming tasks.
The script constructs commands based on user-supplied arguments and invokes them using subprocess.call(cmd). The implementation correctly handles arguments as a list of strings, which is a standard security practice to prevent shell injection vulnerabilities.
[DATA_EXFILTRATION]: The skill is designed to transmit media data to remote network locations using various protocols.
Functionality for RIST, RTP, and Icecast protocols necessitates establishing network connections to transmit streams to external URLs. This is the intended primary purpose of the skill.
[PROMPT_INJECTION]: The skill possesses a potential surface for indirect prompt injection due to the ingestion of external data for command construction.
Ingestion points: User-provided arguments such as --url, --input, and --filter-expr processed by scripts/advproto.py.
Boundary markers: The skill does not implement specific delimiters or instruction-ignore markers when interpolating these arguments into the command string.
Capability inventory: The skill can execute the ffmpeg process and perform network operations via ZeroMQ, RIST, and other streaming protocols.
Sanitization: While argparse provides basic type checking for some flags, string-based inputs like URLs and filter expressions are passed to the subprocess with minimal additional validation, relying on ffmpeg's own parsing logic.
[SAFE]: The skill contains several examples using placeholder credentials for documentation purposes.
The SKILL.md and references/advproto.md files use classic placeholders such as hunter2, password, and secret. These do not represent actual sensitive data exposure.

ffmpeg-rist-zmq