The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by incorporating untrusted user input into a command-line execution.
Ingestion points: User responses to questions regarding 'Project Type' and 'Target Audience' are captured in the variables $PROJECT_TYPE and $INDUSTRY (within the context of $ARGUMENTS).
Boundary markers: There are no delimiters, escaping mechanisms, or 'ignore embedded instructions' warnings applied to these variables before they are used in Step 3.
Capability inventory: The skill utilizes python3 for script execution and mkdir for file system operations in SKILL.md.
Sanitization: There is no evidence of validation or sanitization of the user input before it is interpolated into the bash command execution.
[COMMAND_EXECUTION]: The skill executes a system command using arguments derived from user interaction.
Evidence: The command python3 skills/ui-ux-pro-max/scripts/search.py "$PROJECT_TYPE $INDUSTRY" --domain product -n 3 is executed in Step 3 to validate project categories.
This pattern presents a risk of command injection if the user provides input containing shell metacharacters (e.g., semicolons, backticks) that are not properly escaped by the underlying agent platform.

create