film-demo
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
child_process.spawnSyncto runffmpegandffprobefor video compositing, framing, and metadata analysis. - This behavior is found in
scripts/render.mjs,scripts/compose.mjs,scripts/inspect.mjs, andscripts/verify.mjs. - [DYNAMIC_EXECUTION]: The workflow requires the agent to author a Node.js script (
run.mjs) that utilizes the Playwright library to automate browser interactions, which is then executed locally. - Instructions for this process are detailed in the
SKILL.mdfile. - [EXTERNAL_DOWNLOADS]: Fetches curated wallpaper assets from Unsplash's public download service for video styling purposes.
- This is implemented in
scripts/download-wallpapers.mjs. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data by navigating a browser to arbitrary web applications. This creates a surface for indirect prompt injection if the recorded application contains content designed to manipulate the agent's context or actions during the recording phase.
Audit Metadata