The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates and executes a Python script that uses the subprocess module to call a local command-line tool named zo. This tool is part of the vendor's environment and is used to perform batch processing on file content.
[PROMPT_INJECTION]: The skill contains a surface for indirect prompt injection where untrusted content from an input file is interpolated into a prompt for the zo tool.
Ingestion points: File content is read via the parse_file_lines function in the generated script and passed to the processing loop.
Boundary markers: The prompt construction f"{instruction}\n\nInput: {line.strip()}" lacks delimiters or instructions to the LLM to ignore potentially malicious instructions embedded in the input line.
Capability inventory: The zo command-line tool is described as having full agent capabilities (prompts and tools), which presents a high-impact execution path if instructions are injected.
Sanitization: No sanitization, filtering, or validation is performed on the data read from the input file before interpolation.

for-each-line