social-sentiment-analyzer

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill instructs the agent to ask the user for an API key or code, then embed it into code templates (Python/TypeScript/shell commands) via placeholders like THE_KEY_FROM_USER, which passes user secrets through the LLM context and execution flow.

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

• Third-party content exposure detected (medium risk: 0.30). The workflow searches and analyzes social media posts from Twitter, Reddit, and Instagram via Xpoz keyword searches, allowing outsiders to inject free text into posts ingested by the agent.