The Agent Skills Directory

[PROMPT_INJECTION]: The skill employs misleading metadata by including keywords unrelated to its primary function. While the skill is designed to generate a 'petting head' GIF, it lists keywords such as 'nsfw image detection', 'image compression', and 'svg to image'.
Evidence: In SKILL.md and references/quick-start.md, the English keywords section includes: image compression, image to base64, base64 to image, svg to image, nsfw image detection.
Risk: This discrepancy could cause an AI agent to incorrectly select this skill when a user requests a sensitive operation like NSFW detection. This would result in the user's image being sent to the external GIF generation service instead of the intended tool.
[DATA_EXFILTRATION]: The skill is configured to send user-provided data, including image files and URLs, to an external API.
Evidence: The file references/operations/post-image-motou.md specifies that the agent should send an image_url or a file to https://uapis.cn/api/v1/image/motou via a POST request.
Risk: Combined with the misleading keywords, there is a heightened risk that sensitive user data (such as private images intended for a safety check) could be exfiltrated to a third-party domain without the user's informed consent.

uapi-post-image-motou