motion-graphics-toolkit
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The SKILL.md file includes a utility script snippet using node -e and child_process.execSync for batch rendering. This script interpolates values from a local JSON file (./data/clips.json) directly into a shell command. Because the command uses double quotes for the --props argument, shell metacharacters like $() or backticks within the JSON data will be evaluated by the host shell, allowing for arbitrary command execution.\n- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection through its data-driven rendering capabilities.\n
- Ingestion points: The node -e batch script in SKILL.md reads content from ./data/clips.json to parameterize video rendering.\n
- Boundary markers: The instructions do not define or implement boundary markers (e.g., delimiters) for the data being interpolated into the rendering process.\n
- Capability inventory: The toolkit possesses significant capabilities, including shell command execution (execSync), package installation (npm), and file system access.\n
- Sanitization: While the script uses JSON.stringify to format the properties, it does not sanitize the resulting string for shell-specific escape sequences or command expansion characters.\n- [EXTERNAL_DOWNLOADS]: The skill instructions and audit scripts involve downloading numerous external dependencies from the npm and PyPI registries (e.g., remotion, manim, lottie-web). While these are well-known libraries in the motion graphics domain, the skill relies on these external resources for its core functionality.
Audit Metadata