no-template-ui
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation (README.md) includes instructions for installing required dependencies such as the
playwrightPython package and the@google/design.mdNode.js package. These are legitimate tools for web auditing and design linting. - [COMMAND_EXECUTION]: The skill instructions (SKILL.md) use a Python one-liner to generate random numbers for design variation. Additionally,
scripts/test_probe.pyusessubprocess.runto execute the audit script during a test phase. These are controlled and benign uses of shell execution for internal logic and testing. - [DYNAMIC_EXECUTION]: The auditing tool
scripts/probe.pyuses Playwright'spage.evaluate()to execute hardcoded JavaScript within the browser context. This code is used solely to extract UI measurements (e.g., calculating contrast ratios and identifying element overflows) and does not execute arbitrary external code. - [INDIRECT_PROMPT_INJECTION]: The skill audits external URLs by extracting text and structural data from web pages. This presents a potential surface for untrusted data to enter the agent's context, but the risk is mitigated by structured processing and data truncation.
- Ingestion points: Data is ingested from external web pages via the
probe.pyscript. - Boundary markers: The script returns audit results as structured JSON data, rather than raw text interpolation into the agent's prompt.
- Capability inventory: The skill has access to the local file system (for screenshots) and the network (via Playwright), which are required for its primary purpose.
- Sanitization: The audit logic in
scripts/probe.pytruncates extracted text to 60 characters and focuses on extracting numeric metrics and metadata rather than full content.
Audit Metadata