The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied prompts and transmits them to the external Qwen Image API. The resulting image URL is rendered directly in the agent's response using markdown syntax.
Ingestion points: The --prompt argument in scripts/generate_image.py accepts untrusted user input.
Boundary markers: No delimiters or instructions are used to separate the user prompt from the API request structure.
Capability inventory: The skill performs network requests (requests.post, requests.get) and can write files to the local file system (open(..., 'wb')).
Sanitization: No sanitization or validation is performed on the user-provided prompt or the API-returned URL before rendering.
[COMMAND_EXECUTION]: The generate_image.py script includes a --no-verify-ssl command-line flag. If activated, this disables SSL certificate verification for image downloads, potentially exposing the user to man-in-the-middle (MitM) attacks.
[EXTERNAL_DOWNLOADS]: The skill communicates with dashscope.aliyuncs.com to generate images and downloads the resulting media. As Alibaba Cloud is a well-known service and these operations are central to the skill's functionality, this is documented as expected behavior.

qwen-image