web-design-reviewer

Pass

Audited by Gen Agent Trust Hub on Apr 4, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it directs the agent to ingest content from live, untrusted websites and subsequently modify local source code. A malicious site could potentially include instructions to manipulate the agent's behavior or code modifications.\n
  • Ingestion points: Website snapshots, console messages, and network requests via Playwright MCP tools (scripts/css-risk-audit.py, SKILL.md).\n
  • Boundary markers: Not present; the instructions do not specify any delimiters or warnings to isolate untrusted web content from the agent's core instructions.\n
  • Capability inventory: The skill is intended to perform source code modifications and includes a script that reads local project files.\n
  • Sanitization: No explicit sanitization or validation of the ingested web content is described.\n- [COMMAND_EXECUTION]: The skill provides a Python script (scripts/css-risk-audit.py) designed to scan local project files for risky CSS patterns. The script uses standard libraries (argparse, re, pathlib) and is restricted to specific file extensions, performing read-only operations to identify layout issues like fixed widths or overflow settings.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 4, 2026, 03:39 PM