instagram-slides

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The script includes code that intentionally scans a user's local auth profiles to locate and reuse API keys from other applications (credential harvesting/abuse), representing a high-risk, deliberate behavior.

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

• Third-party content exposure detected (high risk: 0.95). Yes—when --url is used, the script fetches outsider-authored web page text via requests.get(f"https://r.jina.ai/{args.url}"), truncates it into blog_content, and then injects it into the LLM prompt as BLOG CONTENT: {blog_content[:8000]} in plan_carousel().

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

• Potentially malicious external URL detected (high risk: 1.00). The skill issues runtime requests to https://openrouter.ai/api/v1/chat/completions to retrieve an LLM-generated JSON plan that directly controls slide text and image prompts, and it fetches blog content at runtime via https://r.jina.ai/{args.url} which is injected into that LLM prompt.