anydesign

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.85). When the user provides a Website URL, the skill runs WebFetch to retrieve the site’s HTML and then (via scripts/extract_css_vars.py) downloads linked stylesheets/inline <style> blocks; those fetched resources are public web content authored by outsiders, and their CSS/HTML text is ingested into the LLM context for analysis.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)