The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary user-provided text (articles, notes, etc.) to generate video storyboards and scripts. This creates a potential surface for indirect prompt injection.
Ingestion points: External text is saved to capture/extracted/visible-text.txt during Step 1 (Brief).
Boundary markers: None are explicitly defined to isolate user text from the agent's internal instructions.
Capability inventory: The skill can execute local commands (ffmpeg, ffprobe, node) and write files to the project directory.
Sanitization: While no explicit input sanitization is performed, the skill's logic constrains the output to specific video formats and schemas, significantly limiting the impact of any potential injection.
[EXTERNAL_DOWNLOADS]: The skill references the GSAP animation library (v3.14.2) from the jsDelivr CDN in scripts/assemble-index.mjs and scripts/captions.mjs. jsDelivr is a well-known and trusted service, and this usage is standard for web-based content generation.
[COMMAND_EXECUTION]: The workflow relies on executing local scripts and media processing tools. For example, scripts/assemble-index.mjs uses ffprobe and ffmpeg to handle background music looping. These tools are used deterministically for legitimate media tasks. The use of npx hyperframes is consistent with the skill's integration into its parent ecosystem.
[CREDENTIALS_UNSAFE]: The skill includes checks for sign-in status through npx hyperframes auth status to interact with the HeyGen API. It correctly advises users on how to manage their session without hardcoding or leaking credentials in the skill code.

hyperframes-faceless-explainer