The Agent Skills Directory

[COMMAND_EXECUTION]: The serial_mux.py script manages a background multi-plexing service by launching subprocesses. It uses sys.executable to run a server instance of itself and socat to create virtual PTYs for multi-client access. These operations are implemented using argument lists, avoiding shell injection risks.
[COMMAND_EXECUTION]: The serial_mux.py script terminates background service processes (mux and socat) by reading PIDs from a local state file (.embeddedskills/state.json) during management commands.
[COMMAND_EXECUTION]: The serial_log.py script enables the agent to write serial data to a file path specified via the --output argument. This provides a general file-writing capability within the context of its logging functionality.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface.
Ingestion points: Untrusted data enters the agent context through serial_monitor.py, serial_hex.py, and serial_log.py which read directly from serial ports.
Boundary markers: Output is structured as JSON or JSON Lines, though the instructions do not explicitly warn the agent to ignore instructions embedded in the serial stream.
Capability inventory: The skill includes subprocess management in serial_mux.py, file-writing in serial_log.py, and serial data transmission in serial_send.py.
Sanitization: Serial data is decoded using replacement for invalid characters, but no semantic sanitization is applied to the content.

serial