The Agent Skills Directory

EXTERNAL_DOWNLOADS (LOW): The skill installs the 'sharp' library from npm. While 'lovell/sharp' is not on the predefined trusted list, it is a globally recognized, high-reputation open-source project (29k+ stars) with no suspicious installation behavior.
COMMAND_EXECUTION (LOW): The skill performs file system read/write operations and stream processing. These actions are limited to image files and are consistent with the tool's stated purpose of image manipulation.
INDIRECT_PROMPT_INJECTION (LOW): The skill ingests untrusted image data for processing.
Ingestion points: File system paths and streams provided to the sharp() constructor in SKILL.md.
Boundary markers: Not applicable, as the skill does not interpolate data into an LLM prompt.
Capability inventory: The skill can read from and write to the local file system using .toFile() and createWriteStream().
Sanitization: The skill relies on the underlying libvips library to handle binary image data safely. There is no evidence of user-controlled strings being executed as shell commands.

sharp