html-render-check
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses shell scripts (scripts/check.sh, scripts/check-env.sh) to orchestrate the environment setup and validation process. The scripts/validate_html.py script utilizes subprocess.run to invoke the VNU W3C validator using Java, facilitating structural HTML checks.
- [PRIVILEGE_ESCALATION]: Instructions in SKILL.md and the environment verification script scripts/check-env.sh suggest using sudo apt install to install system libraries such as libgbm1 and libxkbcommon0, which are required dependencies for the Chromium browser.
- [DYNAMIC_EXECUTION]: The scripts/render-check.mjs script dynamically loads the Playwright module from a configurable path and executes JavaScript logic (battery.js) within the browser context using page.evaluate(). This allows the skill to perform runtime checks on rendered HTML artifacts.
- [EXTERNAL_DOWNLOADS]: The skill's setup process involves downloading the playwright NPM package and the Chromium headless shell binary from official repositories. These are well-known and standard tools for browser automation.
- [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: The skill processes external HTML files and an optional interactive states JSON file provided as arguments.
- Boundary markers: The skill does not employ specific delimiters or instructions to ignore potential malicious content embedded within the HTML being validated.
- Capability inventory: The tool renders HTML, captures screenshots, and monitors browser console and network activity, reporting these results back to the agent.
- Sanitization: There is no sanitization of the content within the HTML files before rendering, which could allow a malicious file to influence the agent through the reported render verdicts.
Audit Metadata