viz

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to stop and ask the user for the GEMINI_API_KEY before continuing (and the mode uses the Gemini API), which requires the LLM to receive and likely use a sensitive API key without specifying secure handling—an insecure credential-handling pattern.

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

• Third-party content exposure detected (medium risk: 0.30). The skill “viz” acquires attacker-supplied free text/URLs from the user via web_fetch (or web_search snippets on blocks) or from inline pasted text (Step 3), and then pipes that acquired text into mode workflows that call LLM/image/HTML generation tools at runtime (notably references/mode-infographic.md Step 3 and references/mode-visualize.md Step 3 / references/mode-publish.md Step 3), so outsider-authored content can be ingested without selecting a specific trusted item first.