dou-ppt-design
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's setup script (scripts/new_deck.py) automatically installs Python packages python-pptx and lxml using pip. It also executes npm install to set up Node.js dependencies defined in kit/package.json. These are standard package managers used to fetch the skill's operational environment.
- [COMMAND_EXECUTION]: The skill utilizes subprocess.run in scripts/new_deck.py to manage package installations and in kit/build.py to execute node render.mjs. The Node.js script uses Playwright to render slide previews and assets from HTML/SVG templates.
- [DYNAMIC_EXECUTION]: The setup script (scripts/new_deck.py) uses the import() function to dynamically check for the presence of required Python modules (pptx and lxml) before deciding whether to trigger a pip installation.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external user-provided data such as planning documents, memos, and captured images to generate presentation content. This creates a surface where malicious instructions embedded in input data could attempt to influence the agent's logic during the slide creation process.
- Ingestion points: User-provided text documents, planning notes, and images processed by the agent to populate the deck.py definition file.
- Capability inventory: Subprocess execution (pip, npm, node), file system write access (for generating .pptx, .html, and .png files), and browser automation via Playwright.
- Boundary markers: None identified in the instructions to specifically delimit or ignore instructions within the processed user data.
- Sanitization: The rendering engine employs html.escape for the HTML preview generation, but the transformation of user-supplied text into Python code in deck.py relies on the agent's internal safety guardrails.
Audit Metadata