The Agent Skills Directory

[COMMAND_EXECUTION]: The wrapper script meter-cmd.py uses the subprocess module to execute the internal protocol_cli.py tool. This is a legitimate implementation detail used to resolve pathing and interpreter differences between Windows and WSL environments. The script constructs command arguments as a list and does not use shell=True, preventing shell injection vulnerabilities.
Evidence: meter-cmd.py lines 52-54 and 67.
[PROMPT_INJECTION]: The skill creates a surface for indirect prompt injection by ingesting and processing data from external serial devices (electricity meters). If a connected device provides malicious data designed to look like agent instructions, it could potentially influence the agent's behavior when the output is parsed and presented.
Ingestion points: scripts/serial_transport.py reads data from hardware serial ports via the pyserial library.
Boundary markers: Absent; data is presented to the agent in a structured Key=Value format but lacks explicit delimiters to ignore embedded instructions.
Capability inventory: The skill can interact with local hardware serial ports and execute internal Python scripts via subprocess.
Sanitization: Data from the serial port is decoded according to protocol specifications (e.g., BCD, hex, or ASCII) but is not specifically sanitized for natural language instructions.

meter-protocol-serial