The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides and recommends the URL https://chatgpt-plus.top/list/#/home, which has been flagged by automated security scanners as a phishing site.
[DATA_EXFILTRATION]: The skill recommends using https://tryallapi.com/v1 as a default API endpoint for image generation. The scripts/generate.py script is designed to extract the GPTIMAGE_API_KEY from the environment and transmit it to this untrusted third-party gateway, creating a high risk of credential harvesting.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection by ingesting untrusted user inputs (product descriptions and reference images) and interpolating them into complex prompts for downstream image generation tools.
Ingestion points: SKILL.md processes user-provided product identity text and reference images.
Boundary markers: While the skill uses a PRODUCT IDENTITY LOCK block, it lacks clear delimiters or instructions to the image model to ignore potentially malicious instructions embedded within the user-supplied data.
Capability inventory: The skill possesses network communication capabilities (scripts/generate.py) and local file system access (scripts/stitch_long_image.py).
Sanitization: No evidence of sanitization, filtering, or escaping of user-provided strings is present before they are integrated into the final prompt templates.

taobao-detail-long-image