paper-grid
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes shell commands for development and build tasks. Specifically, scripts/paper_grid.py uses subprocess.run to perform a syntax check on generated JavaScript using node --check. Additionally, scripts/fix-frame0.sh invokes ffmpeg to correct rendering artifacts in the output video. These executions use controlled arguments and are standard for video production workflows.\n- [EXTERNAL_DOWNLOADS]: The skill requires playwright and Google Chrome to support its web page capture functionality. These are well-known, trusted tools from reputable organizations (Microsoft and Google). The installation instructions in SKILL.md follow standard practices for setting up a development environment.\n- [DYNAMIC_EXECUTION]: The core library (scripts/paper_grid.py) dynamically generates a JavaScript timeline for GSAP-based animations. The generation logic employs json.dumps for interpolating variable data into the script, which mitigates the risk of code injection. The generated script is validated for syntax at build time but is not executed directly by the Python process.\n- [INDIRECT_PROMPT_INJECTION]: The skill contains functionality to capture screenshots of external websites via scripts/capture_page.py. This represents an ingestion point for untrusted data. Mandatory evidence chain: 1. Ingestion point: URL provided to capture_page.py. 2. Boundary markers: N/A for visual captures. 3. Capability inventory: Subprocess calls to node and ffmpeg, and network access via Playwright. 4. Sanitization: The script only interacts with the page to remove specific UI elements like headers and cookie banners before taking a static screenshot. The risk is minimal as the ingested content is rendered into a static image.
Audit Metadata