The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-uploaded images to create animations, which presents an attack surface where malicious instructions could be embedded in image metadata.
Ingestion points: SKILL.md describes using user-uploaded images as frames or inspiration for the agent.
Boundary markers: The instructions do not specify delimiters or warnings for the agent to ignore embedded instructions in source images.
Capability inventory: The skill documentation in SKILL.md includes the capability to read files and write to the file system using PIL and imageio libraries.
Sanitization: There is no mention of sanitizing input image data or stripping metadata in the provided instructions.
[EXTERNAL_DOWNLOADS]: The skill requires installation of standard Python packages pillow, imageio, and numpy.
Evidence: SKILL.md contains a command to install these libraries via pip. These are well-known, widely-used libraries for image processing and are appropriate for the skill's primary purpose.

slack-gif-creator