The Agent Skills Directory

[External Downloads] (LOW): The skill requires the installation of the google-generativeai package. This is a trusted dependency maintained by the google organization, which is a trusted source. Evidence: pip install google-generativeai in SKILL.md.
[Command Execution] (LOW): The skill suggests using the macOS screencapture utility to generate input files. This is a standard system utility used for the skill's primary function. Evidence: screencapture -i /tmp/bug.png in SKILL.md.
[Indirect Prompt Injection] (LOW): The skill is susceptible to visual prompt injection, where malicious instructions hidden within processed images (OCR or visual reasoning) could attempt to override the agent's behavior.
Ingestion points: Local image files (PNG, JPEG, etc.) passed to the -f flag in the CLI examples.
Boundary markers: Not present in the provided prompt templates.
Capability inventory: The model processes visual input to generate text descriptions, React components, and bug reports.
Sanitization: No input sanitization or validation of image content is performed; the skill relies on the model's internal safety guardrails.

gemini-image