frontend-slides
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: Several scripts in the skill execute shell commands to manage files and trigger system-level actions. Specifically,
scripts/export-pptx.pyusesos.systemto automatically open the generated presentation in the OS's default viewer (openorxdg-open). - [EXTERNAL_DOWNLOADS]: The skill automates the retrieval of various well-known developer libraries and tools from official registries. It installs Python packages like
python-pptx,lxml,playwright, andpillowfrom PyPI, and Node.js tools likevercelandplaywrightfrom NPM. - [REMOTE_CODE_EXECUTION]: The skill downloads and executes remote software as part of its setup and sharing features. This includes the automatic installation of the Chromium browser binary via Playwright and the Vercel CLI tool for hosting slides. It also generates and runs temporary scripts (
export-slides.mjs) to handle specialized rendering tasks. - [PROMPT_INJECTION]: The skill has an indirect prompt injection surface in 'Mode B' (PPT Conversion) and 'Mode C' (Enhancement). It ingests text and content from external PowerPoint files and existing HTML presentations into the agent's context, which could potentially contain adversarial instructions designed to influence the agent's behavior during generation.
- [DATA_EXFILTRATION]: The
scripts/deploy.shscript facilitates the upload of generated slide decks and their associated assets to Vercel, a well-known cloud hosting service. This is a core feature for sharing presentations live and is documented as the intended method for external distribution.
Audit Metadata