The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of curl commands to interact with the EachLabs API at api.eachlabs.ai for fashion-related AI tasks.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data which is then processed by external AI models, creating a vulnerability surface.
Ingestion points: Text prompts, remote image/video URLs, and zip archive URLs specified in the model parameters in SKILL.md and references/MODELS.md.
Boundary markers: No delimiters or instructions are provided to the agent to treat user-supplied prompts as potentially untrusted or to ignore embedded instructions.
Capability inventory: The skill uses network capabilities to send data to the eachlabs.ai API and requests the API to fetch content from user-provided URLs.
Sanitization: There is no evidence of input validation or sanitization for the prompts or URLs before they are passed to the backend API.
[PROMPT_INJECTION]: The references/MODELS.md file defines a disable_safety_checker parameter for several models. Notably, the become-image model sets this to true by default, which effectively disables the vendor's built-in content safety mechanisms for generated imagery.

eachlabs-fashion-ai