The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user descriptions and Replicate API outputs, which are then used in prompts and shell commands.
Ingestion points: Ingests user-provided text descriptions for image generation and tool-generated URLs from Replicate in SKILL.md.
Boundary markers: No explicit delimiters or instructions to the agent to disregard potential instructions within the input data are present.
Capability inventory: The skill utilizes curl for network downloads and writes files to the local public/ directory.
Sanitization: The instructions do not specify sanitization for the user input or the resulting URLs before they are passed to the shell.
[EXTERNAL_DOWNLOADS]: Downloads generated image content from Replicate's well-known service to the local filesystem using curl.

generate-image