The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core workflow of ingesting and processing untrusted data from remote websites.
Ingestion points: Website content is extracted into several files, most notably capture/extracted/visible-text.txt and capture/extracted/asset-descriptions.md, which are then used as source material for narration and visual layout.
Boundary markers: The skill lacks explicit boundary markers or instructions that would tell the agent to ignore or isolate any instructions embedded within the captured website text.
Capability inventory: The agent has the capability to execute shell commands (via npx hyperframes), write project files, and generate executable HTML/JavaScript code for the video compositions.
Sanitization: There is no evidence of sanitization or filtering of the extracted website content before it is used to generate the narration script or composition code.
[COMMAND_EXECUTION]: The skill frequently executes shell commands via npx, relying on the hyperframes CLI tool and tsx for capture, validation, and rendering tasks.
[EXTERNAL_DOWNLOADS]: The process involves downloading arbitrary assets (images, fonts, and text) from user-provided URLs during the initial capture phase.

website-to-hyperframes