The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs and executes shell commands using ImageMagick (magick/convert) and awk for image processing and arithmetic calculations.
The commands interpolate variables such as $IN, $OUT, $STRENGTH, and $CLIP directly into shell strings. For example, magick "$IN" ... "$OUT" and -contrast-stretch ${CLIP}%x${CLIP}% are vulnerable to command injection if these variables are derived from untrusted user input without sanitization.
The procedure utilizes subshell execution for arithmetic: $(awk "BEGIN{print $STRENGTH*100}"). This pattern allows for code execution if the $STRENGTH variable is manipulated to break out of the awk context.
The skill uses magick identify -format "%[fx:standard_deviation]" "$IN", which involves ImageMagick's internal expression language (fx), representing another layer of potential script injection via malformed input.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external files (images) using high-privilege system utilities, creating a potential attack surface for indirect injection.
Ingestion points: The skill reads image files (JPEG, PNG, TIFF, WebP) from the filesystem as specified in the Procedure section.
Boundary markers: No delimiters or instructions are provided to the agent to ignore potentially malicious content or filenames within the processed data.
Capability inventory: The skill utilizes magick, convert, awk, and xargs to perform filesystem operations and image transformations.
Sanitization: There is no evidence of path validation, numeric range checking, or shell escaping for the parameters passed to the ImageMagick and awk commands.

auto-tone