make-video
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external text files, such as briefs and scripts, to drive the design and build phases, which introduces a surface for indirect instructions to influence agent behavior.
- Ingestion points: User-provided script or brief files are read by the Designer and Builder subagents.
- Capability inventory: The Builder subagent writes executable HTML/JS/CSS code, and the skill invokes system commands including npx hyperframes and magick.
- Boundary markers: The subagent prompts lack explicit delimiters or instructions to treat the ingested script content as data only.
- Sanitization: There is no mention of filtering or sanitizing the input script content before it is used to generate plans and code.
- [DYNAMIC_EXECUTION]: The skill generates and executes code dynamically as part of its core video creation process.
- The Builder subagent writes HTML, CSS, and JavaScript to implement visual beats.
- The composition is rendered and inspected at runtime using the hyperframes framework.
- [COMMAND_EXECUTION]: The skill relies on several command-line tools to perform its functions.
- Executes npx hyperframes for inspecting, linting, snapshoting, and rendering compositions.
- Executes magick (ImageMagick) to resize and analyze screenshots for visual validation.
- [EXTERNAL_DOWNLOADS]: The skill uses npx to execute the hyperframes package and its associated tools (inspect, lint, snapshot, render) during the build and critique phases.
Audit Metadata