demo-video
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill identifies and executes a startup command extracted from the local repository's documentation (e.g.,
MakefileorREADME.md). It also executes several generated scripts usingnodeanduv run pythonto orchestrate the video production pipeline. - [EXTERNAL_DOWNLOADS]: The skill downloads voice models (e.g.,
nl_NL-mls_5809-medium) from the official repository of the well-known Piper TTS project to generate narration audio. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from local repository files (README, source code, Makefiles) to plan video scenes and generate mock data for the demo. This represents an attack surface where malicious project content could influence the agent's actions.
- Ingestion points:
SKILL.md(Step 2: Begrijp de app) reads project documentation and source files. - Boundary markers: None present to delimit project data from instructions.
- Capability inventory: The skill can write files, execute subprocesses (
ffmpeg,playwright,node,python), and use theghCLI for network operations. - Sanitization: The skill utilizes predefined scaffolds for script generation, which provides some structural consistency.
- [DYNAMIC_EXECUTION]: The skill generates multiple JavaScript and Python scripts (
record.mjs,record.py,stitch.py,generate_audio.py) at runtime based on its analysis of the project, which are then executed to perform recording and post-processing tasks.
Audit Metadata