meitu-skills

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skills explicitly accept and fetch arbitrary external image URLs and reference images (see meitu-tools mapping of image_url → --image, meitu-beauty/meitu-cutout/meitu-image-adapt Execute steps that download and validate image URLs, and meitu-poster / meitu-carousel Poster Analyse and Execute that use reference images / --image_list to extract style and build prompts), and those untrusted third‑party images are parsed and used to construct prompts and routing/command decisions—so external content can materially influence tool use and agent actions.