The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting instructions and structure from externally pulled style templates and user-provided facts.
Ingestion points: Instruction files located in templates/<slug>/ (including REMIX.md and TEMPLATE.md) and user-supplied research data in references/remix.md.
Boundary markers: The skill mandates two distinct human review steps: "Stop 1 (storyboard)" and "Stop 2 (draft)" to prevent unintended autonomous actions.
Capability inventory: The agent can execute npx hyperframes (authentication, media generation, rendering), use curl for footage downloads, and manage local file writes.
Sanitization: The skill relies on structured order and human review rather than explicit automated sanitization for content interpolation.
[COMMAND_EXECUTION]: The skill utilizes the npx hyperframes and heygen CLI tools for core functionality, including account status checks, template retrieval, and asset generation. It also performs detached background rendering using shell commands (bash -c) to ensure processes persist through long-running jobs.
[EXTERNAL_DOWNLOADS]: The skill uses curl to download media assets from well-known and reputable sources such as Wikimedia Commons, Openverse, NARA (National Archives), and the Library of Congress. These sources are consistent with the skill's purpose of sourcing legitimate stock and archival footage.

style-sampling