distil

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly accepts a --url flag and states "Use WebFetch to analyze an existing site" (Site Specification section and Step 1), and the fetched site content (SITE_SPEC) is passed into subagents and used to drive design generation and file-creation (Step 4), meaning arbitrary public webpages can be read and materially influence agent actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill's runtime pre-flight and resolution steps include npx install commands that would fetch and run remote skill code (e.g., "npx skills add bencium/bencium-claude-code-design-skill@bencium-innovative-ux-designer -g -y"), and those fetched ~/.claude/skills/* SKILL.md files are then read by subagents to drive prompts/instructions, so this is a runtime external dependency that can directly control agent behavior.