The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes external tools and internal scripts to process video. In scripts/render_master.py, subprocess.run is used to invoke ffmpeg and render_hyperframe.py. These calls use argument lists, which is a secure practice to prevent shell injection.
[EXTERNAL_DOWNLOADS]: The skill retrieves assets from several external sources:
Font and styling assets are imported from api.fontshare.com (as seen in shared/_shared.css) and fonts.googleapis.com (as seen in scripts/build_storyboard_preview.py).
Music assets are generated and downloaded via the ElevenLabs Music API (using the fal-ai client) in scripts/fire_music.py.
[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection (Category 8) because it autonomously scrapes content from user-provided brand_url targets.
Ingestion points: Brand homepages, meta descriptions, and Product Detail Pages (PDPs) as described in Phase 0 of the workflow.
Boundary markers: The instructions do not specify the use of delimiters or warnings to ignore embedded instructions in the scraped text.
Capability inventory: The agent has the ability to execute subprocesses (ffmpeg), write files to the local system, and perform network requests to AI APIs.
Sanitization: There is no evidence of sanitization or validation of the scraped text before it is used to generate the ad brief or rendered into video frames. However, the workflow includes human-in-the-loop gates (Gate 1 and Gate 2) which serve as a primary mitigation.
[DYNAMIC_EXECUTION]: The skill uses a dynamic rendering pattern (Category 10) where it generates HTML/JS files (beat-*.html) at runtime and then executes them within a headless browser (Playwright) to capture video frames. This is a standard architecture for the 'Hyperframes' rendering system.

create-value-prop-ad