audio-reply

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly implements "Mode 1: Read URL Content" and example workflows that use WebFetch to fetch arbitrary URLs (e.g., "read it to me https://example.com/article") and extract/summarize that untrusted third-party webpage text to generate TTS, so external content is read and directly influences the assistant's output/behavior and could carry indirect prompt injections.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly WebFetches user-supplied URLs at runtime (e.g., https://example.com/article / https://blog.example.com/new-feature) and injects the fetched text into the assistant's response/TTS, so external content can directly control what the agent says.