html-composition
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes standard command-line and library-based execution for its primary purpose. It employs Playwright's Chromium to capture screenshots and ffmpeg to process video files. The provided Python snippet for Playwright and the ffmpeg filter examples are consistent with the skill's stated goal of visual composition.
- [INDIRECT_PROMPT_INJECTION]: The skill processes HTML/CSS content to render visuals, which acts as a data ingestion surface.
- Ingestion points: Local HTML files (referenced via
file:///paths inSKILL.md). - Boundary markers: None explicitly mentioned for separating instructions from data.
- Capability inventory: Shell command execution for ffmpeg and headless browser execution via Playwright's Chromium (
playwright.sync_api). - Sanitization: None specified; the skill assumes the agent is authoring the HTML or processing trusted layouts.
- [EXTERNAL_DOWNLOADS]: The instructions explicitly discourage external font dependencies to ensure rendering consistency, recommending that fonts be embedded as data URIs or using system-safe font stacks.
Audit Metadata