The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/run.py file uses os.execv to re-execute itself within a dedicated virtual environment if one is detected. This is a standard practice for maintaining dependency consistency and does not pose a security risk in this context as the path is derived from the script's own location.
[INDIRECT_PROMPT_INJECTION]: The skill takes user input, such as brand names and taglines, and interpolates them into prompts for image generation models. This creates a surface for indirect prompt injection where malicious text in the brand name could attempt to influence the image generation parameters. However, the impact is limited to the visual output and is inherent to the skill's primary function.
Ingestion points: User-provided --brand and --tagline arguments.
Boundary markers: The skill instructs the agent to put the brand name in double quotes within the prompt.
Capability inventory: Image generation via internal batch runners.
Sanitization: The skill includes instructions to mask API keys in errors.

logo-maker