The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external URLs as inputs, which constitutes a potential surface for indirect prompt injection via malicious content or metadata in the linked files.
Ingestion points: The source_image_url and target_video_url parameters in the Face Swap workflow defined in SKILL.md.
Boundary markers: No explicit delimiters or instruction-isolation markers are present in the documentation to prevent the agent from being influenced by instructions that might be embedded in media metadata.
Capability inventory: The skill utilizes network capabilities to call the HeyGen API and provides functions for status polling and workflow chaining.
Sanitization: The provided implementation examples do not perform validation or sanitization of the URLs before transmission to the external API. This surface is inherent to the skill's primary purpose of media processing.

faceswap