The Agent Skills Directory

[Prompt Injection] (SAFE): No attempts to override agent behavior or bypass safety guidelines were detected in the instructions.- [Data Exposure & Exfiltration] (SAFE): The skill describes a local process for image analysis. There are no network calls to non-whitelisted domains or attempts to access sensitive system files.- [Indirect Prompt Injection] (LOW): The skill processes external image data which represents a potential injection surface.- Ingestion points: Image files accessed via the disk path parameter.- Boundary markers: The usage example shows prompt interpolation using double quotes, providing basic segmentation.- Capability inventory: The skill calls a local command-line tool describe_image.- Sanitization: No explicit sanitization of the image content or the resulting text description is mentioned.- [Command Execution] (SAFE): The skill references a local CLI tool but does not include scripts that execute arbitrary or dangerous commands. It advises on resource management to prevent GPU overloading.

describe-image