product-launch-video

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). This skill explicitly tells the agent to "relay [auth status] output verbatim" and also checks for environment API keys/credentials (e.g., GEMINI_API_KEY, ~/.heygen), which could cause the LLM to read and echo secret values back, creating an exfiltration risk.

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow can ingest outsider-authored text via Step 1 capture of an arbitrary user-provided URL (or confirmed user-selected site) using npx hyperframes capture "<URL>", then later uses extracted visible text as product truth for narration/story, making indirect prompt-injection possible from poisoned page content.