The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes code patterns that execute external system commands using the subprocess module.
In SKILL.md, the MultiStreamer class uses subprocess.Popen to run ffmpeg. The command arguments are constructed by interpolating variables such as input_source, video_bitrate, and destination URLs derived from StreamDestination objects.
While the use of a list for arguments is a safer practice than shell strings, the lack of explicit input validation in the provided pattern represents a potential surface for command injection if populated with untrusted data.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing external, untrusted data which could influence agent behavior.
Ingestion points: The StreamBot class in SKILL.md implements event_message and handle_commands using the twitchio library, which ingests raw chat messages from Twitch users.
Boundary markers: The provided pattern does not include specific delimiters or instructions to ignore potential commands embedded within the chat text beyond the defined command prefix.
Capability inventory: The skill includes capabilities for network operations (Twitch/YouTube APIs), file streaming (FFmpeg), and software control (OBS WebSocket).
Sanitization: There is no explicit sanitization or filtering logic shown for the incoming chat messages before they are processed by the command handler.

streaming-patterns