kai-video-production
Pass
Audited by Gen Agent Trust Hub on Oct 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill frequently executes shell commands and local scripts for audio/video processing, including 'python tools/voiceover.py', 'python tools/music_gen.py', and 'npm run render'. It also interacts with cloud infrastructure via 'modal app list' and performs environment checks using 'ffmpeg -version'.
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect injection surface by ingesting untrusted data from the project codebase (README.md, MARKETING.md) and external websites via Playwright recording ('/record-demo --url'). There are no documented boundary markers or sanitization steps to prevent embedded instructions in these sources from influencing the generated video scripts or scene plans. Evidence: phase 0 (auto-explore codebase), phase 1 (read MARKETING.md), and Step 2 (browser recording).
- [DYNAMIC_EXECUTION]: The skill generates executable code at runtime by creating Remotion scene components and configuration files (src/Root.tsx, project.json) and then executing them via 'npm run render'. It also uses Playwright automation to drive browser interactions based on AI-generated scripts.
Audit Metadata