The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to ingest and process untrusted external data in the form of user-uploaded images.
Ingestion points: SKILL.md explicitly describes loading user images using Image.open('file.png') for use as direct content or inspiration.
Boundary markers: Absent. There are no instructions or code to isolate or ignore potentially malicious instructions embedded in image metadata or crafted pixel data.
Capability inventory: The skill uses PIL and numpy for image manipulation and references core/gif_builder.py for writing files to the local filesystem. It does not appear to have network or shell execution capabilities.
Sanitization: Absent. No validation or stripping of image metadata (like EXIF tags) is performed before processing.

animated-message-composer