The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data, including product photographs and text descriptions, which are interpolated into prompts for image generation.
- Ingestion points: User-supplied image file paths and product descriptions provided during the creative brief process.
- Boundary markers: The skill utilizes specific, repetitive prompt templates (e.g., "Keep the product itself exactly as it is...") to bound AI behavior and minimize influence from potentially malicious input data.
- Capability inventory: Executes shell commands to discover and run a sibling script from the same vendor.
- Sanitization: Instructions require the agent to visually confirm that the provided image file actually matches the product description before processing.
[COMMAND_EXECUTION]: The skill utilizes shell commands such as bash, curl, and jq to verify environment readiness, query available models from the vendor API (api.ofox.ai), and execute the image generation workflow.
[DYNAMIC_EXECUTION]: The skill dynamically identifies the location of its core execution script (ofox-image.sh) by searching through multiple computed relative and absolute paths (including home directories) and then executes the script using bash.
[EXTERNAL_DOWNLOADS]: The skill references external installation commands for its dependencies from the vendor's repository using npx, though these are documented for user execution rather than automatic agent execution.

product-image