The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external image files, which represents a potential attack surface for instructions embedded in visual data.
Ingestion points: All functions in scripts/see.py accept a path parameter to external image files.
Boundary markers: The skill does not implement delimiters or safety instructions to separate image content from agent instructions.
Capability inventory: The skill has the ability to read local files and write results to the /home/claude directory.
Sanitization: No sanitization or filtering is performed on the image content before processing.

seeing-images