agency-carousel-growth-engine

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). This skill is high-risk: it is explicitly designed to operate autonomously without user confirmation (allowing implicit invocation), scrape arbitrary websites and extract content, and then publish that extracted data publicly via external APIs while self-scheduling persistent executions (cron), which together enable unauthorized data exfiltration and persistent, backdoor-like behavior.

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

• Third-party content exposure detected (high risk: 1.00). The skill's SKILL.md explicitly requires Playwright-based scraping of "any website URL" via analyze-web.js to extract headlines, features, testimonials, competitors, and other site content which the agent then uses to generate prompts, hooks, and publishing decisions—exposing it to arbitrary public web content that could contain malicious or instructional payloads.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly navigates and scrapes arbitrary, user-supplied target website URLs at runtime with Playwright (the "target URL" used by analyze-web.js) and then injects extracted content—CSS colors, copy, testimonials, etc.—directly into the generated Gemini slide prompts, so those external site URLs can directly control the agent's prompts.