The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates downloading registry manifests, component code, and a 33MB machine learning model from the HeyGen HyperFrames GitHub repository and static asset domains. Documented neutrally as these originate from a well-known service vendor.
[DATA_EXFILTRATION]: The hyperframes feedback command sends search queries and feature requests to the vendor's feedback system. Additionally, the hyperframes publish command allows the agent to upload local project files to hyperframes.dev for sharing and review.
[INDIRECT_PROMPT_INJECTION]: The skill processes registry manifests and component metadata fetched from an external source, which represents a potential surface for indirect prompt injection.
Ingestion points: Metadata and code snippets enter the context via the hyperframes catalog and hyperframes add commands.
Boundary markers: No explicit instruction delimiters are used when processing external registry content to distinguish it from system instructions.
Capability inventory: The skill allows for local file creation and modification (hyperframes add), shell command execution, and project rendering.
Sanitization: The instructions do not specify validation or sanitization of the remote registry content before use.
[REMOTE_CODE_EXECUTION]: The skill implements a component installation workflow where JavaScript, HTML, and CSS are fetched from a remote registry and written to the local filesystem for use in projects. This code is later integrated into the project environment and executed during preview and rendering tasks.

hyperframes-registry