The Agent Skills Directory

[PROMPT_INJECTION]: The skill describes a workflow for reverse-engineering video styles by ingesting and analyzing untrusted external content, creating an indirect prompt injection attack surface.\n
Ingestion points: The watch-video and social-fetch tools are used in references/edit-anatomy.md to extract transcripts, on-screen text, and visual frames from external social media URLs.\n
Boundary markers: The instructions do not include specific delimiters or directives to the agent to treat external text as untrusted or to ignore embedded instructions within that content.\n
Capability inventory: The skill has the ability to install software via npm, generate executable code in HTML and React for programmatic video frameworks, and interact with external APIs such as HeyGen.\n
Sanitization: There is no evidence of filtering or sanitizing the extracted text or transcript data before it is incorporated into the agent's decision-making process for creating edit beat sheets.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation and use of external code packages.\n
The instructions provide commands like npm install hyperframes and npx create-video@latest to set up production environments.\n
These downloads target well-known and recognized service providers within the video technology ecosystem.

video