The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user data, including images and natural language instructions, which is a standard surface for indirect prompt injection.
Ingestion points: Natural language messages and image URLs (e.g., image_urls) provided by users to the API endpoint.
Boundary markers: The provided prompt examples use clear instructional patterns and specific constraints (e.g., "Keep the skin texture realistic") to guide AI behavior.
Capability inventory: The skill utilizes network requests to the vendor's API and relies on the each::sense AI model for processing.
Sanitization: The skill instructions emphasize maintaining realism and preserving identity, providing a conceptual framework for safe output, though technical sanitization is handled by the EachLabs platform.
[DATA_EXFILTRATION]: The skill makes network requests to eachsense-agent.core.eachlabs.run and references storage.eachlabs.ai for output delivery.
Evidence: These domains belong to the skill author (eachlabs) and represent the legitimate infrastructure required for the skill's stated purpose of portrait enhancement.

portrait-enhancement