abstract-journey-film

Fail

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches assets and data from various remote sources to facilitate media generation:
  • Downloads Lucide icons from JSDelivr (cdn.jsdelivr.net) for the console interface in scripts/framesmith/console.py.
  • Fetches an SVG icon registry from a public GitHub repository (raw.githubusercontent.com/GLINCKER/thesvg) in scripts/framesmith/media.py.
  • Downloads brand-specific SVGs from thesvg.org in scripts/framesmith/media.py.
  • Communicates with OpenAI's API (api.openai.com) for image generation, text-to-speech, and audio transcription in scripts/framesmith/media.py and scripts/framesmith/narration.py.
  • [COMMAND_EXECUTION]: The skill executes external command-line tools to process media files:
  • Uses ffmpeg and ffprobe for video encoding, audio mixing, and duration detection in scripts/framesmith/timeline.py, scripts/framesmith/narration.py, and scripts/framesmith/motion.py.
  • Spawns a headless browser (chromium, google-chrome) to rasterize SVG assets into PNGs by taking screenshots of dynamically generated local HTML pages in scripts/framesmith/media.py.
  • [DYNAMIC_EXECUTION]: The skill employs dynamic processing techniques for rendering:
  • Parallelizes frame rendering using the multiprocessing pool in scripts/framesmith/timeline.py.
  • Generates temporary HTML wrappers to render SVG icons through a browser engine at runtime.
  • Synthesizes audio tracks programmatically using numpy and the wave module in scripts/main.py.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect instructions through its configuration data:
  • Ingestion points: Reads brief.json and assets.json to define film parameters and image generation prompts in scripts/main.py.
  • Boundary markers: The skill does not employ explicit delimiters or safety instructions when interpolating configuration values into prompts or command arguments.
  • Capability inventory: Significant capabilities including network egress (OpenAI API), file system writes (out/ directory), and subprocess execution (ffmpeg, chromium).
  • Sanitization: Values from brief.json (such as fps and name) are used to drive rendering logic and text overlays without specialized sanitization, though many are cast to specific types (e.g., integers).
Recommendations
  • HIGH: Downloads and executes remote code from: unknown (check file) - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Oct 26, 2026, 10:19 AM