The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently instructs the agent to execute shell commands (e.g., python generate.py --prompt "...") using arguments directly derived from untrusted user input. This pattern is susceptible to command injection if the agent does not properly escape shell-active characters (like semicolons, backticks, or pipes) before execution.
[PROMPT_INJECTION]: The skill facilitates indirect prompt injection by processing untrusted user data that determines the parameters of executable commands.
Ingestion points: User-provided image prompts, style requests, and customization parameters in SKILL.md sections 3.2 and 3.3.
Boundary markers: Absent; user input is directly interpolated into command-line strings without delimiters or instructions to ignore embedded commands.
Capability inventory: Execution of local Python scripts with multiple modes (generate, ultra, inpaint, upscale, remove-bg) via python generate.py in the directories C:\Users\renat\skills\ai-studio-image\ and C:\Users\renat\skills\stability-ai\.
Sanitization: Not present; the skill instructions lack any validation, filtering, or escaping logic for user-supplied strings used in shell interpolation.

image-studio