The Agent Skills Directory

[COMMAND_EXECUTION] (LOW): The skill requires the Bash tool to execute a local processing script (scripts/image_processor.py). This is a legitimate functional requirement for an image processing skill but represents a capability that allows the agent to interact with the underlying shell.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill accepts untrusted text data via the --text parameter and interpolates it into a shell command. This creates a potential injection surface.
Ingestion points: The --text argument used in bash commands within SKILL.md captures user-provided or externally-sourced strings.
Boundary markers: None; the documentation does not specify the use of delimiters or 'ignore' instructions for the text input.
Capability inventory: The skill possesses Bash execution, Read, Write, and Edit permissions, enabling significant file system and command-line actions.
Sanitization: There is no evidence in the documentation of input sanitization or escaping for the shell arguments.

image-processing