The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external image files and handles file system operations, creating a surface for indirect prompt injection.
- Ingestion points: Gimp.file_load in references/ui-and-files.md and Gimp.image_list() in references/complete-examples.md ingest image data into the agent's context.
- Boundary markers: No explicit boundary markers or instructions to disregard embedded data in images are present.
- Capability inventory: The skill uses Gimp.file_save and export_as_jpeg in references/ui-and-files.md for file system writes, and various Gegl operations for data processing.
- Sanitization: No evidence of input validation or sanitization for file paths or image metadata is provided in the examples.
[PRIVILEGE_ESCALATION]: The skill instructs the user to modify file permissions to make scripts executable.
- Evidence: SKILL.md includes the command chmod +x ~/.config/GIMP/3.0/plug-ins/my_plugin.py to enable execution of the developed plugin script.

gimp-plugin