The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided images to generate HTML-based art. This functionality represents an indirect prompt injection surface where adversarial content embedded within a source image could potentially influence the agent if it is subsequently tasked with interpreting the output.
Ingestion points: The process_image function in scripts/ascii_pixel_effect.py ingests image data from local paths.
Boundary markers: None; the skill does not use specific markers to delimit data or warn the agent about instructions within the image.
Capability inventory: The script includes file-writing capabilities via the Python open() function to save the resulting HTML files.
Sanitization: The image is analyzed for luminance and color values to map pixels to ASCII characters; text-based content within the image is not interpreted by the script itself.
[DYNAMIC_EXECUTION]: The generate_html function in scripts/ascii_pixel_effect.py dynamically creates a JavaScript block within the output HTML file to handle canvas animations and interactive ripple effects. This generation follows a static template and is used purely for visual rendering.
[EXTERNAL_DOWNLOADS]: The skill relies on the rembg library, which is documented to download its pre-trained computer vision models from official repositories during the first run. This is a standard and expected behavior for the library's subject detection functionality.

ascii-pixel-art