The Agent Skills Directory

[PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by ingesting untrusted natural language instructions from the user and passing them to ad generation tools.
Ingestion points: The edit_instruction and app_url fields within the generate_app_ad_creative tool call and REST API examples in SKILL.md accept arbitrary user input.
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the embedded edit_instruction as untrusted data or to ignore potential injection attempts within that text.
Capability inventory: The skill utilizes network-capable tools including generate_app_ad_creative, get_app_ad_creative_job, and meta_ads_upload_ad_image_from_url to interact with external advertisement platforms.
Sanitization: There is no evidence of input validation, sanitization, or filtering applied to the user-supplied instructions before they are interpolated into the API workflow.

ad-creative-edit