valet

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

• Third-party content exposure detected (high risk: 1.00). The skill's "Designing a New Agent" workflow explicitly fetches and parses external URLs (e.g., github.com SKILL.md via raw.githubusercontent.com, skills.sh listings, npmjs.com/PyPI pages) using WebFetch/WebSearch to import user-provided third-party content and incorporate it into agent design, so untrusted web content is read and can influence subsequent tool use and decisions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly runs a remote installer with /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)" (executes remote code at runtime) and also instructs fetching raw GitHub content (e.g. raw.githubusercontent.com or github.com/.../SKILL.md) at runtime to inject instructions/skill files into the agent, both of which meet the criteria for risky external control.