The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell command templates involving magick, xargs, and awk. These templates rely on variable interpolation (e.g., $IN, $OUT, $STRENGTH, $CLIP) which can lead to command injection if the agent does not properly escape or sanitize these inputs before they are executed in a shell environment.
[COMMAND_EXECUTION]: The calculation step awk "BEGIN{print $STRENGTH*100}" involves dynamic evaluation of the $STRENGTH variable. This allows for arbitrary code execution within the awk process if the variable is manipulated to contain malicious code.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection through its processing of external data.
Ingestion points: The skill accepts files and directories as input, which are processed by the agent.
Boundary markers: There are no markers or instructions provided to the agent to treat filenames or file metadata as untrusted data.
Capability inventory: The skill utilizes powerful system tools including magick, xargs, and awk via subprocess execution.
Sanitization: No sanitization, validation, or escaping logic is defined for the input parameters or file paths before they are used in commands.

auto-tone